The work list

The DNA proves Derrick descends from Henry Robert Wear and Martha Maud Castle as a couple — 253 of the 297.5 cM he shares with his closest match lie outside every region he shares with any Castle cousin. The couple had two sons, and in June 1948 one was thirty-one and the other was eleven. What is missing is a record naming Leonard as anybody's father. Two death notices were requested from the Master at Cape Town on 15 September; see the DNA page.

Gillian Agnes Louisa Nel, born Wear, and Patricia Mary Maud Grant, born Wear. Each file holds a death notice with a field for the father's full names. Sent to Mr D Keyster, the estate controller for file numbers ending in 1. His auto-reply diverts only appointment queries, so a copy request should stay with him. Ms N Mphetshwa has left the section — that cc was wasted, and on 15 September her own auto-reply CONFIRMED it in her own words: "kindly note that I am no longer working at the Deceased Estate Appointment section". Mr Keyster's auto-reply also fired and diverts APPOINTMENT queries only, to Ms Portia Arendse — so a copy request correctly stays with him. Both confirmations, neither an answer. VINDICATED 15 September by row 33: six marriage registers across thirty years and two form designs have been read and NONE records a parent. The J294 death notice asks for the father's full names and the mother's maiden name, which makes this letter the best South African route to the one fact the archive cannot document — not a redundant one. ANSWERED 15 September at 09:57 — the first substantive reply on this row. Mr Keyster asks for a formal application on an attached form, APPLICATION FORM FOR COPIES.pdf, with proof of payment, submitted 'on groundfloor' at 45 Castle Street, Dullah Omar Building, Cape Town. That last part is an IN-PERSON instruction and David is in Australia — but Mr Keyster answered by email, so email submission may well be acceptable and must be asked rather than assumed. His reply also gives neither the fee nor the banking details, and nothing can be paid without them. A reply is drafted asking exactly those two things. The form itself is a PDF attachment this archive cannot open, so David has to read it. Replied the same hour, Gmail 1a0a482ea47212a0, asking the two questions and — deliberately — restating that David is not a relative and claims no interest in either estate, so that if his office limits copies to interested parties the refusal arrives before the money does.

WLD 2545/1950 — IVONE versus KURT, restitution of conjugal rights. She sued him. That is the step a deserted spouse takes, and it sits between the divorce he filed in 1948 (WLD 1950/1948) and the one he filed in 1951 (497/1951 and 2368/1951). Four actions in three years, in both directions, with Derrick born in the middle of them. It hid because NAAIRS spells both names wrong — Luwenski, Figaretto — so every sweep missed it twice over. Sent to Enquiries2@dsac.gov.za, with 2545/1950 named as the one to do first.

This would test the whole Letchworth reconstruction against a South African record rather than an English index. It cannot be emailed: Ms Mavis Xaba of the Reading Room confirmed on 14 September that MHG files are bound and cannot be scanned. Added on 15 September to the commission the Lerena/Booyzen session already placed with a private researcher — same depot, same visit, and flagged as the least important of the seven.

Both marriages are recorded on the civil registers as entered into WITH an antenuptial contract, so the deeds exist. They sit in a Deeds Registry, which is a different institution from the archives and will not act on a letter or a telephone call. Stuck for weeks; on 15 September the researcher was asked whether she does that work or can name someone who does. AMENDED 15 September: there are THREE, not two. The Wear-Dunlop contract of 30 October 1951 was found the same evening by reading certificate 1397 at full resolution, and it is the only one of the three that names Leonard. Request 18 carries the correction and is held until the researcher replies.

Three went into threads opened in June 2025. The fourth — to the keeper of the seven-person tree that is the only source giving Leonard a mother called Castle — was the first cold approach this archive has made, and it tells a stranger that her father probably had a son she has never heard of. It also went to an account belonging to a woman who died in March 2024, so it may be read by her family or by nobody.

Their terms forbid republishing without written permission and this archive is on the open web. Submitted 15 September. If refused, those six facts come off the public pages into the working notes and the site will say a source asked for them to go, rather than deleting them quietly. It costs the argument nothing — the Wear reconstruction rests on English indexes and the Government Gazette, neither of which is theirs.

Two acknowledged: the Arquivo Histórico Militar in nine minutes, and Bydgoszcz with a warning that a complicated query can take over a month. Everything else is silence. See the status page. ONE OF THE ELEVEN HAS ANSWERED, 17 September, and it is KOSZALIN. Reference O-S.6342.406.2026, dated 16 September, signed by KRZYSZTOF SKRZYPIEC, Head of the Slupsk Branch; the case officer is Krzysztof Chochul. IT IS A NEGATIVE AND A WELL-MADE ONE. The archive holds NO registers of the former German civil registry office at CZARNE — Standesamt HAMMERSTEIN, Kreis Schlochau — FOR 1874 TO 1899, and NO BOOKS OF THE JEWISH COMMUNITY AT CZARNE AT ALL. It adds that they hold no information about what became of either series or where it might now be, which is a more honest answer than a bare nil return. WHY IT MATTERS: HAMMERSTEIN IS GEORG LEIBHOLZ'S BIRTHPLACE. He was born there on 7 June 1884 per Gedenkbuch en1100801, and an 1884 birth register is the one document that would name his parents and settle whether he is a Leibholz of this family at all. THIS ARCHIVE'S POSITION IS THAT HE IS NOT — Jacob was born at SCHUBIN in POSEN on 8 August 1884, two months later and in another province, and two men cannot be brothers on those dates. THAT FINDING NOW RESTS ON THE GEDENKBUCH ALONE, because the record that could have overturned it does not exist. And SALLY LEIBHOLZ, born 1847, is before civil registration began, so only the synagogue books could have held him — and those are the ones that are gone entirely. TEN LETTERS STILL UNANSWERED. THE REPLY ASKS FOR AN ACKNOWLEDGEMENT to slupsk@koszalin.archiwa.gov.pl, which was drafted at requests/28-SEND-THIS-to-koszalin.txt and SENT 17 SEPTEMBER, message 1a0ac71a500bc77f, inside their own thread, to slupsk@ as asked and copied to the case officer Krzysztof Chochul. ONE SMALL THING WORTH WRITING DOWN AFTER THE dac.gov.za BUSINESS: the letter's own footer gives the domain as koszalin.ap.gov.pl while the mail comes from koszalin.archiwa.gov.pl. Both appear live; the acknowledgement uses the address they asked for.

Done 15 September. Read a third time, and this time from the archive's own microfilm rather than from a photograph in the family tree: digital film 008109682, frame 2047, image ark 3:1:3Q9M-CSJ1-SJQK. FamilySearch's own pages would not render — the session had been rate-limited to about one response every thirty-five seconds, which the site's front end shows as a hang — so the page was rebuilt locally from its 108 deep-zoom tiles and every cell magnified between four and five times, with the sideways columns rotated upright. One real correction: the couple's address is 108 DUNOTTAR Street, SYDENHAM, not Dunstan Street, Braamfontein. Everything the argument rests on holds, including the one that matters — column (10) reads With A.N.C. at 5.2x, four letters, not Without.

Worked hard on 15 September and not settled — but the evidence is now correctly weighed, which it was not before. Three routes failed. NAAIRS is down a second day, returning "The connection to the Search Manager Server is inactive" with an empty database list. Gazettes.Africa returns nothing for "Leonard Aubrey Robert Wear", "Wear, Leonard" or "L. A. R. Wear" in any year — and that proves nothing, because the control damns it: a search for Luwinski returns eight documents and not one is Kurt's own 1969 estate notice, which certainly existed. FamilySearch has both his marriages and no death record at all. What did change: GenDatabase's own profile page says in its own words that this record is INCOMPLETE as per their verification, and prints his date of birth as the corrupt string 0017. So 1979 is a damaged entry in a state index and 1973 is a daughter's family tree, and neither is a document. The archive should stop preferring 1979. The cheap way to close it is GenDatabase's single Verified Report at R20, which releases the identity number, the place of death and the full date — a purchase, so it is David's to make and not this archive's. NARROWED 15 September: he was alive on 13 SEPTEMBER 1969, buying his father's grave at West Park. That rules out nothing — both 1973 and 1979 are after it — but it is a hard floor, from a municipal record. || SHARPENED DECISIVELY 16 September. Four instruments have now looked: the NAAIRS estate index (South Africa to end 1976) has no Wear estate 1970-1976; the Johannesburg cemetery registers hold six Wears and none is him; the GRO death index for England and Wales 1970-1985, >99% complete and printing DATES OF BIRTH, holds 241 Wears and none is him; and the GRO consular registers structurally cannot contain a South African death of either year. SO THERE IS EXACTLY ONE PLACE IN THE WORLD WHERE THIS DEATH CAN BE RECORDED: THE SOUTH AFRICAN DEATH REGISTER AT THE DEPARTMENT OF HOME AFFAIRS — and GenDatabase's 1979 entry is a transcription of precisely that register. That changes its standing entirely. It is not one disputed source among several; IT IS THE ONLY SOURCE THAT COULD EXIST, and the family tree's 1973 has now failed three separate tests and has nothing left to stand on. The archive still cannot SAY he died in 1979 — a corrupt transcription is a corrupt transcription — but it now knows exactly what to ask for and of whom: ONE DEATH-REGISTER ENTRY AT HOME AFFAIRS, PRETORIA. 17 September: THE GAZETTE ROUTE HAS BEEN TRIED AND IT DOES NOT ANSWER. All 38 documents carrying both WEAR and AUBREY were read and none is his estate. An estate notice is the one free document that would print his date of death. The negative is filed with its limits attached and they are real — phrase search is broken on this corpus (the control, 'Kurt Israel Luwinski', returns nothing although the man is in the archive eight times), and an estate under the threshold is never advertised at all. THE ROUTE THAT IS LEFT IS THE FAMILYSEARCH JOHANNESBURG CEMETERY SET, 1840 to 2019, which covers West Park — and that is waiting on a sign-in, not on a discovery. A FIFTH INSTRUMENT LOOKED, 17 SEPTEMBER, AND THE REASON IT FOUND NOTHING IS NOW KNOWN RATHER THAN GUESSED. FamilySearch's South African civil death registration is a NAME INDEX TO 1966 - exact-year counts 61,899 for 1960, 74,131 for 1965, 59,263 for 1966, then 178 for 1967 and nothing after. KURT DIED IN 1968, HENRY ROBERT IN 1969, LEONARD IN 1973 OR 1979. All three fall after the last year this index covers, so their absence from it says nothing whatever about them. AND THE ONE SET THAT WOULD HAVE ANSWERED IT WAS OPENED AND MEASURED: "South Africa, Transvaal, Estate Files, 1950-1993" is, in its own words, IMAGES OF DEATH NOTICES - the document that names a dead man's parents - and its title brackets both candidate years. ITS WAYPOINTS SAY IT IS PUBLISHED FOR 1955 TO 1960 ONLY. Seven waypoints, and the Orange Free State equivalent has fifty-seven running to 2006, so this is a publishing gap and not a survival gap. RE-CHECK IT WHENEVER FAMILYSEARCH ADDS TO IT: it is the one instrument that would settle row 96 without Home Affairs, and it would name Leonard's parents in the same stroke.

Not reachable from the sources available, and kept here so nobody tries it again the same way. A son of Bruce born in the 1960s would be in his sixties now, so he is not in a death database; and South African civil marriages on FamilySearch stop at 1974, so a marriage in the 1980s or 1990s is not there either. GenDatabase cannot find living people. The Y-test route needs contact with living Wears, not records — which makes it a question for the people already written to, not a search. Retested 15 September at the family's push, by the one route left — Michelle Ann Wear, b.1978, d.2017 Johannesburg, whose estate would name her parents. She is in no gazette and not in GenDatabase at all. Two routes, both closed.

Answered: they were gone before 29 September 1939. The 1939 Register has seven Wears in Hertfordshire and the household is not among them, and a national search returns Henry Robert Wear in the 1891, 1901, 1911 and 1921 censuses and then stops. The absence is real rather than redaction — born 1889 and 1895, he and Martha would be 137 and 131, so their records would be open if they existed. One daughter stayed: Jeannett (E M) Wear, 1919, 18 Ridge Avenue, Letchworth, aged twenty on register night. She followed later and died in South Africa in 2001.

Run, and the useful answer is a negative. Ivone is in no South African death record under any name she used — not Salkow, not Luwinski, not de Witt, not Figueiredo, and no Ivone among the six women in the verified deaths carrying that maiden name. She married three times in that country and died out of its reach. That is what dying in Porto looks like from this side, and it is the first corroboration this archive has for a death it has never been able to date.

A purchase, and this archive does not spend other people's money. It would place him on a branch young enough for a surname to mean something. Of the 7,748 public members of the R-U106 project only 34 sit under R-S11477, whose common ancestor lived around 200 BCE — so the surnames on that branch are ninety generations away and mean nothing at all. SHARPENED 15 September by row 45. The argument for the Big Y is no longer 'more detail'. The classic haplotree shows forty-two branches below R-S11477, and its two immediate children are R-Z27559 and R-BY42072 — and R-BY42072 is TAGGED ENGLAND, as is R-FT370170 below it (13 testers, c. 550 CE). The Y evidence has so far only been able to say NOT ASHKENAZI. A Big Y that landed Derrick on one of those two branches would make it say something positive and geographically pointed for the first time.

Raised twice and overruled twice, which is the family's decision and not this archive's to make — but it is recorded here because it is still live. Five living people have now been written to about a finding that concerns a man who has not, so far as this archive knows, been told that his family is writing to them.

The cause of the duplicate letter. Projects/SHARED-CONTACTS.md now sits in both this archive's notes and the Lerena/Booyzen one, listing every shared contact and its live state, every address known to be dead and why, and the constraints that apply to both. One rule at the top: before writing to any institution, search the sent mail.

The tracker recorded what was sent and never what was said back, so Ms Xaba's reply — that MHG files are bound and cannot be emailed — sat unread for a day while this archive went on framing every estate request as just send me the one page. There is now a What has actually come back table on the status page.

Founded on a mistake, and kept here rather than deleted. A full day went into Edwina as Derrick's possible half-sister. She was born Edwina Ann GREEN and married Bruce Wallace Wear — Leonard's younger brother — in 1980. A Wear by marriage, no blood relation at all. Every dead end around her had been the answer saying so: her parents were unfindable under Wear because they were Greens, and she was still a Wear at 69 because it was her married name.

Founded on a mistake. A week went into the surname Halsey because it was the name printed on a DNA match card. It was the wrong half of a married couple: the 1952 register gives the birthplaces as Cape Province for him and ENGLAND for her, and the English link was the wife. There was never a Halsey marriage to find, which is exactly why the England and Wales index came back empty.

Standing work, and the reason this archive is worth anything. Published corrections to date: the Castles were Hertfordshire, not Norfolk; Caroline was a Wallis, not a Walls, and the tree was right where this archive was wrong; Derrick's birth year leaked onto the DNA page and was removed; the generation claim was walked back within the hour; and Henrique was called the son of a marriage that the 1948 register proves never happened. BROUGHT UP TO DATE 17 SEPTEMBER. The research log had stopped at 15 September while the archive went on making mistakes. NINE ENTRIES ADDED, of which SEVEN ARE MARKED WITHDRAWN OR CORRECTED: reading a dead hostname as a dead service; publishing two archive-index zeroes that were three records each; claiming DNA above George William Castle on a surname without opening the tree; having the emigration backwards, with the father leaving first; LEAKING THREE LIVING MATCH HOLDERS INTO A BUILD BECAUSE THE GUARD THAT CATCHES IT WAS NEVER WIRED INTO THE BUILD CHAIN; the Martha Giddings quarters; and the WALLIS/Walls contradiction that had sat on a page for a fortnight. The two genuine finds of the same days — Caroline James and Pix Road — sit in the same list in the same voice, which is the point of keeping it. 20 SEPTEMBER, AND THE COUNT FOR ONE DAY IS SIX. Six times today this archive raised as new something it already held: the Chevrah Kadisha (source of correction 1); correction 5's geography argument, which the next panel already made; Willi's 1946 questionnaire, re-read hours after being written up; Siegfried's birth date, published as something "nothing else did" give when Mapping the Lives had it that morning; WALTER of Levetzowstraße, whose absence from the census was this file's own omission; and row 202, which proposed writing to Poland for the Hammerstein registers that were asked for on 14 September and answered on the 16th. THE CAUSE IS THE SAME EVERY TIME AND IT IS NOT MEMORY. It is writing a row or a paragraph off the back of a fresh find WITHOUT FIRST READING THE FILES THAT WOULD ALREADY ANSWER IT — corrections.json, letters.json, the requests directory, and the data file being appended to. THE RULE THAT FOLLOWS: before a new row proposes writing to anybody, grep letters.json and requests/ for the place name. Before a data file claims a fact is new, grep data/ for it. Both are one command. || WIDENED 20 SEPTEMBER, SECOND TIME THE SAME DAY. The rule was "read letters.json and requests/ before drafting a request". IT IS NOW: BEFORE WRITING UP ANY FIND, GREP THE SITE FOR THE NAME. A gazettes.africa sweep produced "Paula Luwinski, a commercial traveller, Mark Hotel" and this archive wrote it up as a discovery about Kurt's widow. The Kurt page already had both notices, already identified her, AND ALREADY CARRIED A PUBLISHED CORRECTION OF THAT EXACT MISREADING — the commercial traveller at the Mark Hotel is Phillip Crystal, the man being sequestrated, and Pauline is the creditor who brought the application. The archive repeated its own corrected error from the same snippet, four days later. || WIDENED AGAIN, 20 SEPTEMBER, AND THIS TIME THE RULE ITSELF WAS THE PROBLEM. The rule said "read letters.json and requests/ before drafting". BOTH WERE READ. letters.json recorded that a letter went to the Landesarchiv Berlin on 14 September about "Kurt and the Leibholz brothers in Berlin", and requests/ held no copy of it — because THIS ARCHIVE NEVER KEPT THE TEXT OF WHAT IT SENT, only a one-line summary of it. So a supplement was drafted that re-explained who everyone was and re-asked for Randvermerke the original letter had already asked for in the same words. THE FIX IS STRUCTURAL, NOT BEHAVIOURAL: every sent letter's full text now goes into requests/sent/ at the moment it is sent, and the letters.json summary records WHAT WAS ASKED, not just the subject. The 14 September letter has been recovered from the mail thread and saved. A one-line summary of an outgoing letter is not a record of it. || AND ONE OF A DIFFERENT KIND, 20 September, late: A COMMIT WAS PUSHED ON A RED BUILD. The command chained the build to the commit with a semicolon instead of &&, so check-evidence reported "FAIL — 26 problems; evidence is not reaching the page" and git committed and pushed anyway. Two clean rebuilds immediately afterwards pass, so the failure was a mid-build race — evidence.json regenerated after the pages were rendered on that one pass — and the pushed state is sound. THE FAULT IS NOT THE RACE, IT IS THE SEMICOLON. This archive's own standing convention is to chain with && precisely so that a failing gate stops what follows, and the convention was broken in the one place it exists to protect. Build, then commit, as separate steps or joined by &&, never by ;. 11 OCTOBER — TWO MORE, AND ONE OF THEM IS A ROW CORRECTING ITSELF. 9500917: row 260 said FORTY-FOUR of the fifty-two Gedenkbuch Leibholz were Berlin and it is FORTY-SEVEN — five plain-Berlin entries, not three, three with two boroughs, not two, and one reached Berlin by way of Paderborn. The row's own figures never summed to its own total, 43 against a stated 44, so the arithmetic was wrong before the data was. 9500918: Wilhelm of Weinstraße 6 was demoted from Kaufmann to Vertreter in the 1939 volume and not, as this archive had it, only from Vertreter to Bauarbeiter in 1941 — THE FIRST STEP CARRIES NO “ISRAEL” BESIDE IT, which is the part worth keeping. AND A THIRD THING THAT IS NOT A CORRECTION BUT BELONGS HERE: ROW 260 SAT BLOCKED FOR NINETEEN DAYS ON PAGE IMAGES THAT HAD BEEN READ TWO DAYS BEFORE IT WAS BLOCKED. Nothing was published wrongly; a task was simply asleep inside a note that said it was waiting. The only reason it was found is that the images were read a second time. 11 OCTOBER — A SEVENTH REDISCOVERY, AND IT IS A ROW'S FAULT AND NOT A READER'S. Row 203 was worked today on the premise that the Spier household Jacob Leibholz lodged with was unidentified. IT HAS BEEN IDENTIFIED ON THE LEIBHOLZ PAGE SINCE 29 SEPTEMBER — the 1939 Teil IV entry in full, the owner read as Gertrud Eisenberg née Haake and Spier H. as Hans Spier, born 2 December 1894. The work was not wasted (it added the 1940-1943 volumes, which are the actual finding) but HALF AN HOUR WENT ON ANSWERING AN ANSWERED QUESTION. THE CAUSE IS STRUCTURAL AND IT IS WORTH NAMING: A WORK LIST ROW IS A RECORD OF WHAT SOMEBODY ONCE WANTED, NOT A RECORD OF WHAT IS KNOWN. Nothing in this archive reads a row against the page that may already have answered it, and the rows are long enough that a later paragraph can answer an earlier one inside the same row. On 20 September the count for one day was six; this is the seventh over three weeks and the first where the answer was sitting on a published page rather than in a data file. AND THE CORRECTION ABOVE WAS WRONG TOO, WITHIN THE HOUR. It said the Leibholz page had read the 1939 Treuchtlingerstraße entry and that the other four volumes were new. THE 1941 ENTRY IS ALSO ON THAT PAGE, quoted in full since 21 September, owner's “SARA” and all. So the rediscovery was larger than the first admission of it, AND THE SECOND READING OF THE PAGE FOUND IT WHERE THE FIRST HAD NOT — because the first stopped at the panel that matched the search term and the answer was in the panel after it. THE REAL YIELD IS THREE VOLUMES, NOT FIVE, AND TWO OF THEM ANSWER A QUESTION THE PAGE POSED IN WRITING: “the 1942 and 1943 volumes would close it further … those two pages have to be found by eye.” A ROW IS NOT A RECORD OF WHAT IS KNOWN; NEITHER, IT TURNS OUT, IS A SINGLE GREP OF THE PAGE THAT KNOWS IT.

Founded on two birth years and killed by a third date. This archive proposed Henry George Charles Halsey (1906–1993) and Mary Ellen Halsey (d.1981) as Mervin's parents because they were the right ages. But he married Evelyn Ruth in 1932 and Mervin was born on 15 June 1930 — two years earlier; Mary Ellen has no spouse recorded and is not his wife; and Mervin was born in the Cape Province, which nothing places any of them in. Off the critical path regardless: the Halsey line stopped mattering the moment column four of the 1952 register was read.

Drafted as request 15 and DELIBERATELY NOT SENT. The plan was a letter to Home Affairs. Searching the sent mail first — the rule written after the Xaba duplication — turned up a reply from a professional South African researcher in May 2020 saying that certificates of any kind, abridged, unabridged or vault, can only be applied for IN PERSON at Home Affairs, that third-party applications are no longer accepted, and that an applicant living overseas must go through the South African Embassy or Consulate in their country. So the letter as conceived would have gone to an address that cannot act on it — the fourth dead address of this project, avoided this time before sending rather than after a bounce. It is redrafted to the High Commission in Canberra and asks HOW rather than asking for the record. And it surfaces something uncomfortable: if eligibility really is limited to immediate family, then for the 1948 marriage that means the two spouses, both long dead, and their son of record. This document cannot be got without Derrick, which collides with row 15. Raised once, here, and left with David.

Struck 15 September, with an answer rather than a shrug. The film viewer loaded once the throttle eased, and the frames either side of ours were opened: frame 2046 is certificate 764, Louw and Swanepoel; frame 2048 is certificate 295, Mowbray and Gemmer. Both are certificate FRONTS, both 19 June 1948. The film runs front, front, front across its 3,228 images. So the reverse of the sheet was never photographed, and no amount of patience with FamilySearch will produce it. The question is not dropped — it is folded into row 22, because the letter to Home Affairs about entry 858 has to ask for both sides of it now.

Found tonight by reading the register of Leonard's second marriage at full resolution: certificate 1397 of 1951, column 10, reads "with". So a notarial ante-nuptial contract was executed for Leonard Aubrey Robert Wear and Marjorie Eileen Dunlop, born Walker, and registered in the Johannesburg Deeds Office. This is the third ANC on the archive's list, after Kurt's two, and it is the only one of the three that names LEONARD. An ANC recites each party's full particulars and is not confined to the register's eleven columns, which makes it the best document about him this archive has ever had a route to. Goes to the Pretoria researcher with rows 4 and 5.

New tonight, from the 1940 marriage register: Leonard's occupation is entered as ARMY and his residence as ROBERTS HEIGHTS, PRETORIA — the Union Defence Force's principal base — on 7 September 1940, one year and one day after South Africa entered the war. Second World War service records are held by the SANDF Documentation Centre in Pretoria, and an attestation form gives full name, date and place of birth, next of kin and in most cases THE NAMES OF THE PARENTS. That is exactly what the archive is missing. The entire chain rests on Leonard being a son of Henry Robert Wear and Martha Maud Castle, and the only source saying so is a seven-person tree built by his daughter. An attestation form would turn that into a government record. SENT as request 16. Three things settled first. This mailbox already has an unanswered enquiry from David, sent 9 September for the sibling archive about the Booyzen family, so the draft opens by naming it — two cold letters in six days is how the Xaba duplication happened. The address the whole internet gives, sandfdoc@mweb.co.za, is DEAD: it bounced on 9 September with Relay access denied, and the Centre has moved from Visagie Street to Irene; archive@dod.mil.za is the live one. And he was almost certainly NOT an officer — officers were gazetted by full name, the control searches return their entries, and Leonard returns nothing under six spellings — so the letter asks for an attestation and a medal-roll entry rather than a commission. SENT 15 September to archive@dod.mil.za, Gmail 1a0a3b70c6389154. UNBLOCKED 16 SEPTEMBER: THERE IS NOW A SERVICE NUMBER TO QUOTE. 141272, Corporal, South Africa Union Defence Forces, from the War Office prisoner-of-war lists WO 392/15 and WO 392/21. This row has sat blocked because SANDF cannot search on a name and a guess. A NUMBER IS EXACTLY WHAT THEY CAN SEARCH ON. The request should also say he was a PRISONER OF WAR in Germany, listed 1943-45, because repatriation and POW files are held separately from ordinary service files and asking for both by name saves a second letter. A South African service file gives date and place of birth, next of kin, and the addresses given on enlistment and discharge — WHICH BEARS DIRECTLY ON THE 1948 QUESTION, because it would place him, or fail to, in the year that matters.

The civil registers have no parents' columns. Catholic sacramental registers do. Leonard and Phyllis were married at Rosebank by W. P. Whelan, who filled in the Denomination line himself as "Catholic Priest", and they were married BY BANNS rather than on a licence — so the intention was read in church on three Sundays and generates its own entry as well. A parish marriage register of 1940 ordinarily names the father and mother of both parties. This is a church record, so it sits entirely outside the Home Affairs eligibility problem in request 15. Identify the Rosebank parish in the Archdiocese of Johannesburg and write to it. SENT as request 17. The parish is the Church of the Immaculate Conception, 16 Keyes Avenue, Rosebank, established 1936 — four years before the wedding. Not asserted as certain: the civil register gives only the word Rosebank, so the letter asks to be redirected if the marriage belongs to another parish or to a chaplaincy at the Roberts Heights camp. Both letters disclose plainly that David is not a descendant and makes no claim to be. SENT 15 September to Info@rosebankcatholicchurch.co.za, Gmail 1a0a3b72c965ecf1. ANSWERED 22 SEPTEMBER 2026 — IN A WEEK, COURTEOUSLY, AND WITH A DOCUMENT THIS ARCHIVE ALREADY HAD. They sent the ORIGINAL MARRIAGE REGISTER, the Union civil form B.M.D. 3 no. 13 of 1940, completed by Father Whelan as marriage officer. IT HAS NO PARENTS' COLUMNS, and every field on it — Army, Roberts Heights, the banns, both witnesses — has been on the DNA page since 17 September, rebuilt from FamilySearch film 008161359 frame 475. THE FIRST WRITE-UP OF THIS REPLY ANNOUNCED THE OCCUPATION AND THE ADDRESS AS NEW FINDINGS AND WAS WRONG; a document arriving by email feels new, and the envelope was the only new thing. WHAT IS REAL: the parish ANSWERS, which matters more than the scan, because the books that name parents are still in their building. And a clean image now exists — published at /wear-robertson-1940.png — where the site previously had only a table of fields. One discrepancy is flagged and NOT silently fixed: the DNA page reads the first witness as MARY UNA ROBERTSON, the parish scan looks like MARY INA ROBERTSON, and two hard readings of one word prove nothing. THE ROW STAYS OPEN ON THIS ARCHIVE: a second letter is owed, thanking them and asking by name for the LIBER MATRIMONIORUM and the BANNS REGISTER, since the difference between the parish's own books and the state form it fills in is not obvious from outside. CLOSED 22 SEPTEMBER 2026 BY THE PARISH ITSELF. Arlene at Rosebank answered all five questions. The 1940 register SURVIVES, CONTAINS THE ENTRY, and DOES NOT NAME THE PARENTS — “Unfortunately not, as attached”. There is NO separate sacramental register and NO banns book: “the Catholic Church keeps one Marriage Register, the banns are included in the attached”. Fr Whelan confirmed at Rosebank from 1935; the books are at the parish, not the Archdiocese. ROUTE 2 TO LEONARD'S PARENTAGE IS CLOSED, and the DNA page now says so. A pure thank-you was sent the same day with no further request in it.

THE OBVIOUS DOCUMENT THIS ARCHIVE HAS NEVER ORDERED, and it should be embarrassing that it took until now. Every route tried this week has been an attempt to get Leonard's parents out of a SOUTH AFRICAN record — marriage registers with no parents' columns, an estate that does not exist, a military file, a parish book. His ENGLISH BIRTH CERTIFICATE names his father and mother outright. The index entry is already held: WEAR, Leonard A R, Hitchin, Q1 1917. A GRO certificate costs about GBP 12.50 and arrives as a PDF in days. If it names Henry Robert Wear and Martha Maud Castle, the single weakest link in the whole chain — that Leonard is their son, currently resting on a seven-person family tree built by his daughter — becomes a government record. A purchase, so David's, but this is the cheapest decisive document on the entire list. || REFERENCE PINNED 16 September from FreeBMD, so the certificate cannot be ordered wrong: HITCHIN, March quarter 1917, vol 3a page 1244 — WEAR, Leonard A R, mother CASTLE. The birth certificate names the father and gives his occupation and the family's address at the time.

The companion to the row above, and the generation behind it. The archive established that the ONLY Wear-Castle marriage in the whole English index is Henry R Wear to Martha L Castle, Hitchin Q4 1914. The certificate names BOTH FATHERS, with their occupations. That does two things at once: it confirms the couple from a state record rather than an inference, and it pushes the line back a generation on both sides at a stroke. Same cost, same speed. || REFERENCE PINNED 16 September from FreeBMD, so the certificate cannot be ordered wrong: HITCHIN, December quarter 1914, vol 3a page 1552 — WEAR, Henry R, spouse CASTLE. The 1914 certificate names the fathers of BOTH parties — Henry Robert Wear's father and Martha Maud Castle's father on one sheet — so it pushes this line back a generation in a single document.

DONE — and the row should never have been written. It claimed the sibship had been cited repeatedly while only four of the six were named. That was false: section 7 of the Wear notes names all six with quarter and district, from FindMyPast's England & Wales Births searched on mother's maiden name CASTLE — six results, all Hitchin. Leonard A R Q1 1917, Janette E M Q4 1919, Kathleen E Q1 1921, Cynthia Y E Q2 1931, Linda A Q2 1933, Bruce W Q3 1936. Closed with the reason attached rather than deleted: the work list exists to stop the archive re-doing work, and a row written without re-reading the notes does the opposite.

DONE, and it produced the most important thing on the list. The second son was already named — BRUCE WALLACE WEAR, Q3 1936 Hitchin, in South Africa by 1959, four marriages and three divorces in the Witwatersrand. What was NOT known is whether he is dead. The notes said "Bruce was gone by 2014", inferred from his wife Edwina's estate notice naming no surviving spouse. Tested against two indexes on 15 September: the Verified Death Database holds Edwina and NOT Bruce; the Government Gazette estate notices hold Edwina and NOT Bruce. And the inference has an obvious alternative — he had already divorced three wives, so a fourth divorce leaves her with no surviving spouse while he lives. Absence is not proof of life and he could have died abroad. But the honest position is now 'not recorded as dead in South Africa', not 'gone by 2014'. See row 31: born 1936 he would be ninety, and he is not a cousin — he is LEONARD'S BROTHER, a direct male-line son of Henry Robert Wear.

SHARPENED 15 September, and it now has a name. The candidate is not a distant cousin to be hunted: it is BRUCE WALLACE WEAR, Leonard's younger brother, born Q3 1936, who carries the same Y chromosome the archive is trying to match against Derrick's R-S11477. Row 30 establishes that he is not recorded as dead in South Africa. If he is living he is ninety. ONE Y-37 test from him confirms or destroys this entire investigation in a single reading, for a fraction of the Big Y in row 14. He is a living person: nothing about him goes on the site beyond the index entry already published, and THIS ARCHIVE WILL NOT APPROACH HIM. Whether anybody does is David's, and it walks straight into row 15. THREE NAMED CANDIDATES, 17 SEPTEMBER, and they come from the 1881 census rather than from hope. Robert Wear had three brothers in his father's house at Jamaica Road in 1871 — WILLIAM b.1850, JAMES b.1852 and HENRY b.1854, all three harness makers at their father's trade. JAMES IS STILL IN JAMAICA ROAD TEN YEARS LATER: RG11 piece 574, folio 56, page 8, schedule 501, 42 JAMAICA ROAD, BERMONDSEY — James Wear, 29, harness maker, born WALWORTH; wife Ann; Gertrude 7; AND TWO SONS, LOUIS b.1878 AND HARRY b.1881. A SECOND HOUSEHOLD a few streets off at 2 FENNER ROAD, RG11/567/128/60 schedule 1238, has WILLIAM WEAR, 38, SADDLER — his father's exact trade, the same parish, the same sub-district — with a son WILLIAM b.1875. THAT ONE IS PROBABLE AND NOT PROVED AND THE FILE SAYS SO: the age gives 1843, which is five to seven years out against both the 1851 and the 1871 returns, and the birthplace column is blank. SO THE Y-CARRYING CANDIDATES ARE LOUIS WEAR b.1878, HARRY WEAR b.1881 and — IF THE FENNER ROAD MAN IS OURS — WILLIAM WEAR b.1875. Three boys, all Bermondsey, all in the male line from William Wear the saddler of Walworth. AND THIS IS THE ONLY ROUTE LEFT ON THIS SIDE. Henry Robert's own siblings cannot do it: LOUISA'S ONLY CHILD DIED IN THE QUARTER HE WAS BORN, Rosetta never married, and the two Hull half-siblings share Emma Hawkins but not Robert Wear. NEXT: follow Louis and Harry to 1891, 1901, 1911, 1921 and 1939, then to marriages and sons; and search the match list for the surname WEAR against the twelve-region Wear signature, which a Jamaica Road cousin should hit and a Castle cousin should not. FOLLOWED THE SAME AFTERNOON, AND THE TRAIL BREAKS — SAID PLAINLY RATHER THAN LEFT HOPEFUL. The death index for ST OLAVE SOUTHWARK gives the next chapter: ROBERT WEAR b.1859 died Q1 1890, and WILLIAM JOHN WEAR b.1849 DIED Q2 1890 — TWO OF THE BROTHERS INSIDE TWO QUARTERS, at thirty-one and forty-one, and the second entry hands us a middle name the censuses never gave. ANN MARIA WEAR b.1852 died Q1 1895; two Wear children, ALBERT GEORGE b.1886 and PHOEBE b.1890, died in 1891. BUT NEITHER LOUIS NOR HARRY IS IN THE 1911 CENSUS OR THE 1939 REGISTER under that name, and the 1891 census shows why the search is hard rather than why they are gone: EIGHTEEN WEARS IN BERMONDSEY IN 1891, and their birth counties are MIDDLESEX AND LANCASHIRE, not Surrey. Even the James Wear of 1891 was born in Middlesex; ours was born at Walworth. BERMONDSEY IS TO WEAR WHAT BALDOCK IS TO CASTLE. The brothers cannot be followed on a name from here, and the next step is not another index search — it is the ST OLAVE SOUTHWARK BAPTISM AND BURIAL REGISTERS, which name parents on every line, or the 1891 household pages read one by one for a Surrey-born James.

DONE 15 September. Certificate 146 of 1953, Johannesburg, frame 1087, image 3:1:3Q9M-CSV4-R9JG-K, rebuilt from 336 tiles. Morris Isaac Salkow, 31, born Transvaal, DIVORCEE, salesman. Ivone Luwinski formerly de Witt born de Figueredo, 25, Portuguese East Africa, DIVORCEE, SALESLADY, of 9 Rudney Road, Rossmore. Without ante-nuptial contract; special licence 220116; and ORDER FILED against BOTH names — which is the second document, and the first she signed, proving the 1951 divorce from Kurt was actually granted. Salkow turns out to have married three times, 1943, 1953 and 1960, so this marriage ended too. AND IT CORRECTED THIS ARCHIVE THE SAME DAY: the age panel published on /1948/ that morning read the February and June 1948 ages as a two-year jump in four months. With the 1953 figure the real shape is a stated age drifting steadily YOUNGER — 22, then 24, then 25 where 1924 requires 28 or 29. The panel is rewritten, and the rule it now carries is that no inference in this archive may rest on an age Ivone gave a clerk.

ANSWERED, and the answer is NO. The South African civil marriage register never asked for parents — not in any decade and not on either form. The archive had read five (1940, 1948, 1951, 1952 eleven-column; 1956 on the fourteen-field B.M.D. 3(a)). A genuine 1970 register was pulled and read to test whether a later redesign added them: image 3:1:3QS7-L9Z5-GQPP, entry no. 2 of 1970, form B.M.D. 3. Its fourteen fields are names, date, solemnized at, identity number, race, age and date of birth, ante-nuptial contract, personal status, occupation, place of birth, address, banns or licence, consent, remarks. Not one is a parent. The state asked more and more about the parties themselves and never once asked whose children they were.

STRUCK by row 33 the same evening it was proposed. It existed only on the hope that a later marriage form named parents, which would have made Gillian's 1969 register do for free what the Master of the High Court is being asked to do. The 1970 form has fourteen fields and no parents, so these three registers will not name Leonard and there is no point ordering them for that purpose.

DONE 15 September, and it paid. The sheet had been opened once, for column 4, which redirected the whole archive; nobody went back for the rest. Certificate 51 of 1952, frame 2149, image 3:1:3Q9M-CSJY-YGDD. Three things came out. (1) 64 GREENFIELD ROAD, GREENSIDE, JOHANNESBURG — Cynthia's address at 21 and a spinster, so very probably the household address of Henry Robert Wear and Martha Maud Castle, and the first address this archive has ever had for them. (2) Banns, at St Martin's-in-the-Veld, denomination ANGLICAN CHURCH — so the Wears were Anglican, and the 1940 Catholic ceremony came from Phyllis's side, which is what request 17 guessed and now has evidence for. (3) Column 10 reads WITH, a FOURTH ante-nuptial contract. And one disappointment worth recording: column 9, Consent given by, reads 'Own' for both — had Cynthia been a minor it would have carried her father's name outright.

DONE 15 September by driving NAAIRS itself rather than a mirror of it. THE SWEEP FOUND NO WEAR ESTATE IN THE TRANSVAAL BETWEEN 1970 AND 1976; the last is the father's, MHG 10245/69, remarks SURVIVING SPOUSE MARTHA MAUD WEAR. Martha Maud's own 1986 estate is NOT absent — it is out of range, because the deceased-estate index runs to the end of 1976 and stops (2,246 estate records in 1976, one in 1977, zero after). That has to be got from the Master's office direct, which is row 2's correspondence. The measurement is the real result: it makes Leonard's two published death years asymmetric for the first time — 1973 is INSIDE the index and he is not in it; 1979 is three years past its last year and cannot be tested at all. Published on /leonard/ with the limits stated.

DONE 15 September, and closed for a structural reason worth stating. Every exact form returns zero — "Wear, Leonard Aubrey Robert", "WEAR, L. A. R.", "L A R Wear" — and every loose form is useless: Wear+engineer+Johannesburg returns 5,568 documents and Wear+Fairlands+Johannesburg returns 3,469, because WEAR IS AN ORDINARY ENGLISH WORD appearing in trade names (More Wear Industries, Fashion Wear, Boys' Wear Manufacturing), wool-labelling prosecutions and railway reports about flange wear. THIS IS THE EXACT OPPOSITE OF THE LUWINSKI CASE AND THE ASYMMETRY MATTERS: Luwinski is rare, returns eight hits in the whole archive, all read, so a null means something. Wear returns thousands, none of them people, so a null means nothing. The archive will stop treating the two nulls as the same kind of object.

DONE 15 September, both halves, and the plain answer is YES THE LISTS EXIST AND NO KURT IS NOT IN THEM. The Government Gazette publishes naturalisation in two series — intention-to-apply notices under section 10 of the South African Citizenship Act 1949, and lists of certificates actually issued; 246 gazettes carry the heading and both were read in full for 1975 and 1976 while working row 58. A full-text search of every South African gazette returns LUWINSKI exactly eight times and this archive had already read all eight — six Piccadilly Hat Company partnership notices, the 1972 Paula application and the 1975 Pallou's Dress Centre notice. NONE IS A NATURALISATION NOTICE. Luwinski is a rare surname and the same search demonstrably finds him, so unlike the WEAR null this one is worth something. NAAIRS cannot help: a search of the central-government database for NATURALIZATION returns 342 documents and every one is a POLICY FILE OR A BILL — the 1926 Act, its 1927 amendment, South West Africa bills. Individual naturalisation files are not indexed at item level anywhere this archive can reach, so the remaining route is a letter, which is row 61.

DONE 15 September, and it returned a coincidence that has to be handled rather than filed. TAB WLD 155/1915 and 329/1919: DIVORCE, MAUD MABEL LONGHURST (BORN CASTLE) VERSUS AUBREY LEONARD — the index writes the defendant by forenames alone because the surname is shared, and TAB MHG 3632/50 gives him in full as AUBREY LEONARD LONGHURST, died 1950, second wife Athalie born van Aardt. SO: A WOMAN CALLED MAUD, BORN CASTLE, MARRIED TO A MAN CALLED AUBREY LEONARD, IN JOHANNESBURG, TWO YEARS BEFORE LEONARD AUBREY ROBERT WEAR WAS BORN TO MARTHA MAUD CASTLE. Three elements of this archive's central identity in one line. It is probably nothing: Martha Maud was nineteen in 1915, married in Hertfordshire since 1914 and on the 1921 census at Letchworth, so she is not the plaintiff, and both Castle and Maud are common. BUT IT WOULD ANSWER A QUESTION NOBODY HERE HAS ASKED OUT LOUD — WHY JOHANNESBURG? Emigrating families go where somebody already is, and if Maud Mabel Castle was Martha Maud's aunt or elder cousin then the Castle family had a Johannesburg branch twenty years before the Wears arrived, and a boy christened Leonard Aubrey in 1917 was named after his mother's relative's husband. NOTHING IS CLAIMED. Written up in castle-baldock.md with the caution attached, because a later reader who found these three references without it could do real damage. Also swept: WEAR+CASTLE returns zero across all repositories — no South African record joins the two names.

WORKED and came back empty, which is itself worth having. Searched the FamilySearch collection that holds both his parents — South Africa, Gauteng, Johannesburg, Cemetery Records 1840-2019 — plus every South African death, burial and Find a Grave collection. LEONARD AUBREY ROBERT WEAR APPEARS EXACTLY TWICE IN ALL OF FAMILYSEARCH'S SOUTH AFRICAN HOLDINGS: his 1940 marriage and his 1951 marriage. No death, no burial, no grave. That is a real negative rather than a gap in the index, because the same collections hold Henry Robert at West Park in 1969, Martha Maud in 1986, Edwina in 2014 and Michelle in 2017. The likeliest reading is that he did not die in Johannesburg — he was divorced twice and could have gone anywhere. Which points straight back at row 10: GenDatabase's PLACE OF DEATH field is paywalled, and the R20 report would say which province to search. Blocked on that purchase, which is David's. RE-RUN AND WIDENED 17 SEPTEMBER, and it is now a much stronger negative. The earlier run searched the cemetery collection and "every South African death, burial and Find a Grave collection". This one enumerated ALL THIRTY-TWO South African collections FamilySearch holds and searched the surname EXACTLY across all of them: 111 records, every one read. Leonard is in two of them, both marriages. He was then searched under WARE, WEIR, WIER, WEARE, WEARNE and WEARS, in case the negative was a spelling. It is not. The reading stands and the reason is now measured: see row 10.

WORKED to the end of what can be done from here, and the answer is that he cannot be placed. Three addresses in fifty-eight years: 72 Pix Road Letchworth in 1921, Roberts Heights Pretoria in September 1940, Polana Court Abel Road Berea in October 1951. BETWEEN 1941 AND 1950 THERE IS NOTHING AT ALL — no address, one recorded fact, that a daughter was born in 1941 — and 1948 sits in the middle of that hole. Nothing places Leonard in Johannesburg in 1948 except that he married there in 1940 and again in 1951, which is an inference from either side of a gap. The gazette route is now exhausted: full-text searches for Polana Court, Dunottar Street, Abel Road and Greenfield Road against the surname return nothing, and the only two hits for Greenfield Road and Wear are advertisements for ladies' wear. What is left — street and telephone directories, the Deeds Office property index, the voters rolls — all need somebody physically in Johannesburg or Pretoria, and none of them costs a fee worth arguing about. Added to the held brief for the researcher. Published on /leonard/ with the silences given the same weight as the entries. SUPERSEDED IN PART the same evening: a fourth address and a fourth fixed point — 271 Wilson Street, Fairlands, 13 September 1969, from the West Park grave register, folio W 8. Leonard was alive and in Johannesburg on that date, which is the last certain sighting the archive has of him.

Row 12 is marked done, and it is done only in the sense that the emigration is BOUNDED — after the 1921 census, before 29 September 1939. The UK outward passenger lists, BT27, would give the actual ship and the actual date. Two notes have carried this as a to-do for days without anyone touching it. It also answers something the archive has never asked: did Leonard travel WITH his parents, or separately, and at what age. || BLOCKED 16 September: THE FAMILYSEARCH SESSION HAS SIGNED OUT and this row cannot move without it. The catalogue and persona APIs now return 401 and the site redirects to the login page. This archive will not enter credentials on David's behalf. One sign-in in the browser unblocks this row, rows 42, 60 and 88 together, and the post-1970 marriage register test in row 81. || NARROWED 16 September WITHOUT A PASSENGER LIST, and closed at that. NO WEAR MARRIED A FOOT IN ENGLAND AND WALES BETWEEN 1938 AND 1955 — and GenDatabase has Kathleen Edith Wear marrying James Cecil Foot in 1945, so SHE WAS IN SOUTH AFRICA BY THEN. Put with the rest, the window closes hard: Q3 1936 Bruce born at Hitchin, family still in England; 29 September 1939 the household ABSENT from the 1939 Register; Q4 1939 Janette marries Edward Bailey AT HITCHIN, having stayed for the wedding; 7 September 1940 Leonard marries at Rosebank, Johannesburg; 1945 Kathleen marries, not in England. THE HOUSEHOLD LEFT BETWEEN AUTUMN 1936 AND SEPTEMBER 1939, AND TWO OF THE SIX CHILDREN FOLLOWED SEPARATELY. That is a three-year window fixed by records rather than inferred from silence, and it is as far as this goes without the outward passenger lists, which are behind a paywall.

DONE 15 September, by ratio rather than by segments, and the limits are published with it. For each shared match, cM-with-Derrick divided by cM-with-Penelope-Browne measures how much of that person's relationship to Derrick is left unexplained by the documented Castle line. All ten shared matches ranked. THE CLOSEST MATCH 3.62, THE 96 cM MATCH 3.32, a 92 cM match 1.96, a 70 cM match 0.97, then the Castle-side core at 0.10 to 0.42. The 96 cM match is in neither of the tables the archive built on 14 September: 96.0 cM with Derrick and only 28.9 with the Castle descendant. On the reasoning that found the Wear line, that is what a Wear-side cousin looks like. Caveats stated on the page: the ratio conflates 'not on Browne's line' with 'distant from Browne', and the denominator is inflated at the bottom because those five are Browne's own first cousins. The bottom of the table is nearly meaningless; the top is not.

DONE 16 September — not by getting the 96 cM match's eight segments, which are still behind the chromosome browser's Advanced panel, BUT BY BUILDING THE THING THAT WILL TEST THEM IN SECONDS WHEN THEY ARRIVE. The archive's argument has always been made by EXCLUSION: these regions are the ones the Castle cousins do NOT share. That is almost impossible to test a new match against, because it asks the reader to hold twelve absences in mind. So it was computed the other way round from the local segment data and written down as coordinates. Subtracting every region either documented Castle-side cousin shares, from the closest match's twelve segments, leaves 253.2 cM IN 12 NAMED REGIONS, now published on /dna/ and stored as data/wear-signature.json. THE COMPUTATION VALIDATES ITSELF: it reproduces this archive's previously published 44.3 cM shared and 253.2 cM not shared TO THE DECIMAL, from an independent method. It also produced the Castle side positively for the first time — the two Castle cousins overlap each other by about 151 cM across nine regions on chromosomes 1, 2, 5, 6, 12 and 18, which is what a tight documented pair should look like, while the closest match overlaps them by only 44. STATED LIMIT: the subtraction is on physical coordinates with cM apportioned by length, and cM and base pairs do not scale together, so the last digit should not be over-read.

STANDING CHECK, and the date is the point. Checked 14 September, re-checked 15 September: R-U106 unchanged at 34 of 7,748, and STILL NO WEAR, WEIR OR WARE anywhere on the branch — the same five surname projects as yesterday (Ballew 16/62, Stafford 8/96, Blakely 8/68, Byrd 5/102, York 4/85). A dated negative is worth more than an undated one. BUT the re-check found a project the first pass missed: R1b-P89.2, where 33 OF 62 PUBLIC MEMBERS SIT ON DERRICK'S EXACT BRANCH — 53 per cent, against 0.4 per cent in R-U106. If a surname ever surfaces from the Y side it will surface there, and this archive was watching the wrong project. Re-check both, monthly, and record the date each time. RE-CHECKED 20 SEPTEMBER, and the date is still the point. R-U106 is now 7,755 public members, up seven from 7,748 on 15 September, and THE COUNT ON DERRICK'S BRANCH IS UNCHANGED AT 34. R1b-P89.2 unchanged at 33 of 62. The five surname projects are identical to a week ago — Ballew 16/62, Stafford 8/96, Blakely 8/68, Byrd 5/102, York 4/85 — and THERE IS STILL NO WEAR, WEIR OR WARE ANYWHERE ON THE BRANCH. Two geographical projects are on the list that this row had never written down: ENGLAND GB GROUPS EIJ (NORMAN), 5 of 5,073, and the GERMAN LANGUAGE AREA project, 4 of 5,405. Neither changes the reading; the first quietly supports it, since R-S11477 is a North Sea lineage and an England/Norman geographical project is where that should show. Next re-check due about 20 October. RE-CHECKED 22 SEPTEMBER 2026, AND A WEEK HAS CHANGED NOTHING. R-U106 now shows 7,755 public Y-DNA members with 34 on R-S11477 — the project has grown by seven since 15 September and THE BRANCH HAS NOT MOVED. R1b-P89.2 is unchanged at 33 of 62, still the densest concentration anywhere. The five surname projects are unchanged to the member: BALLEW 16/62, STAFFORD 8/96, BLAKELY 8/68, BYRD 5/102, YORK/YORKE 4/85. AND STILL NO WEAR, NO WEIR, NO WARE — tested this time by matching the whole rendered page rather than by reading down it, which is the difference between a check and an impression. TWO PROJECTS THIS ROW HAD NEVER RECORDED, AND THEY BEAR ON THE ARGUMENT: the branch also appears in ENGLAND GB GROUPS EIJ (NORMAN), 5 of 5,073, and in the GERMAN LANGUAGE AREA DNA RESEARCH PROJECT, 4 of 5,405. Both are geographic rather than surname projects, so neither can produce a name — but this site argues that R-S11477 is NORTH SEA AND NOT ASHKENAZI, and an English-Norman project and a German-language project are precisely where that argument expects to find it. Also on the branch: R1b ALL Subclades 21/23,680, R1b French 2/2,462, R1b1a2 Iberian 1/2,090, and R1b Basal Subclades 0/2,150. RE-CHECKED 10 OCTOBER, MECHANICALLY, AND THE EARLIER PASSES WERE SEEING A THIRD OF IT. The Discover page is an Angular shell over a PUBLIC JSON FILE — discover.familytreedna.com/resources/y-dna/R-S11477.json, 226 KB, no account — so the check is now a command and not an afternoon. HEADLINE UNCHANGED AND DATED A FOURTH TIME: no WEAR, WEIR or WARE anywhere in the whole 226 KB record, not just in the project list. R1b-U106 is 34 OF 7,785, against 34 of 7,755 on 20 September — THE PROJECT GREW BY THIRTY AND THE BRANCH DID NOT GROW AT ALL. R1b-P89.2 is 33 of 63. The five surname projects are identical to three weeks ago. BUT THE RECORD LISTS THIRTEEN PROJECTS AND THIS ARCHIVE HAD ONLY EVER RECORDED FIVE. The eight never written down include two GEOGRAPHIC ones that are worth more than the surname ones — ENGLAND GB GROUPS EIJ (NORMAN) 5 on the branch, and GERMANY-YDNA 4 — and neither has ever been looked at. The seventeen countries were never recorded either: Unknown Origin 167, United States 30, England 18, Germany 7. In data/ydna-s11477-2026-10-10.tsv. 11 OCTOBER — THE ARITHMETIC WAS DONE AND IT REVERSES THIS ROW. The two geographic projects were the thing this row told the next session to look at. THEY CARRY NO SIGNAL. Against a base rate — 21 of 23,730 in the R1b ALL Subclades master project, 0.0885 per cent, which is what “a project contains this branch” looks like when nothing is going on — ENGLAND GB GROUPS EIJ is 5 of 5,088, 1.1 TIMES BASE, and GERMANY-YDNA is 4 of 5,420, 0.8 TIMES BASE. The branch sits in a large English project and a large German project AT CHANCE. The instruction to go and read their member lists is withdrawn: it would cost an afternoon and return nine men whose presence means nothing. AND THE FIVE SURNAME PROJECTS, WHICH THIS ROW CALLED WORTH LESS, ARE ENRICHED BETWEEN FIFTY-THREE AND TWO HUNDRED AND NINETY TIMES: BALLEW 16 of 62, 292x · BLAKELY 8 of 68, 133x · STAFFORD 8 of 96, 94x · BYRD 5 of 102, 55x · YORK 4 of 85, 53x. R1b-P89.2's 592x is CIRCULAR — it is a project defined by a SNP on this branch. WHICH FORCES A CORRECTION TO THIS ROW'S OWN SENTENCE. “No surname has surfaced and none is predicted to” is TRUE IN ITS MEANING — nothing names Derrick's biological father's line, and that is not weakened — and MISLEADING IN ITS WORDS, because five surnames sit on this branch far above chance. The honest form: FIVE SURNAMES SIT ON THIS BRANCH AT FIFTY-THREE TO TWO HUNDRED AND NINETY TIMES CHANCE AND NOT ONE OF THEM IS WEAR, WEIR OR WARE. It does not make him a Ballew; it names where a paper trail would have to come from if one ever comes. THE COUNTRIES FLATTEN THE SAME WAY: raw counts read Anglo-German, but against each country's own pool the United States is 2.3x, England 1.5x and GERMANY 0.95x — exactly chance — and ONE HUNDRED AND SIXTY-SEVEN OF 242 BRANCH MEMBERS STATE NO ORIGIN AT ALL, so every rate is computed on under a third of the branch. The scope codes were in the data all along: H haplogroup, L surname, G geographic, which confirms “two are geographic” from the record rather than from the project names. WHAT TO WATCH INSTEAD: the five surname projects' branch counts as a time series, and a SIXTH surname project appearing in the list. In data/ydna-s11477-enrichment.tsv.

DONE 15 September. /leonard/ is live and in the menu. His life as far as it is documented, with every row tagged documented, inferred or disputed; the case that he is Derrick's biological father; and THE CASE AGAINST SET OUT AT THE SAME LENGTH, which is the point of the page. Row 50 is folded into it as a section rather than scattered elsewhere.

The 44-against-253 argument is the strongest thing in this archive and it is currently three paragraphs of prose. It is a picture: chromosomes along an axis, Derrick's shared segments with the closest match in one colour, the Castle cousins' in another, and the overlap visible or not visible at a glance. Anyone who sees it understands the argument in five seconds. Anyone who reads the prose has to take it on trust. DONE 15 September: the chromosome map is live on /dna/, computed at build time from the segment coordinates rather than typed in, so the 297.5 / 44.3 / 253.2 figures in the prose are derived from the data beside them. Matches labelled by role only; the copy in site/src/data has the surname keys stripped.

DONE 15 September, and the old rule is replaced rather than reaffirmed. All twenty-six addresses this project has used were re-resolved and audited against the six that have actually failed. MX CAUGHT ONE OF THE SIX. The reason is now understood and published: dac.gov.za and dsac.gov.za resolve to the SAME mail gateway, so a renamed department keeps a perfectly valid MX record and the rejection happens inside, where DNS cannot see it. A live MX proves a mail server answers for the domain and nothing about the department, the office or the person. The replacement procedure, ordered by what actually catches things: search the sent mail first for prior contact AND prior bounces (three catches so far); prefer an address the institution itself gave you over a directory entry; check the institution still exists under that name and at that address; MX last, for a mistyped subdomain; then watch a day for a bounce. Written into SHARED-CONTACTS.md in both archives and published on /letters/.

DONE 15 September. /letters/ is live and in the menu, built from site/src/data/letters.json and RECOMPUTED ON EVERY BUILD, so the dates are live rather than written down once and forgotten. Fifteen letters, each with an interval chosen for the recipient rather than one global number: 10 days for a commissioned researcher who has been asked for rates, 21 for a Master's office or a national archive or a parish, 28 for an overseas archive in another language. Sorted by soonest due. The page says plainly why it exists — this archive had already sent one correspondent two emails in four hours and another a duplicate enquiry it had forgotten making.

DONE 15 September, as a section of the new Leonard page rather than a page of its own. Six things are named that would make this archive abandon the identification — a birth certificate naming different parents, a male-line Wear Y test that does not match R-S11477, an attestation form with a different next of kin, a third son in the birth index, a J294 naming a different father, and evidence he was out of the country in 1948. Written down BEFORE the documents arrive and published, so it cannot be quietly revised afterwards.

ANSWERED 15 September, and now established across THREE separate films rather than assumed from one. Film 008109682 (the 1948 register): frames 2046 and 2048 are certificates 764 and 295, both fronts. Film 008161359 (the 1940 register, frame 473, certificate 13): frame 474 is an Afrikaans certificate from Turffontein, number 206 — a front. Film 008019258 (the 1951 register, frame 2700, certificate 1397): frame 2701 is certificate 1400, Siegbert Buchholz of Germany — a front. THE REVERSES OF THESE SHEETS WERE NEVER PHOTOGRAPHED. It is systematic to the collection's filming, not a quirk of one reel, and no amount of patience with FamilySearch will produce them. That strengthens request 15: whatever is written on the back of any of these registers survives only in the original bound books at Home Affairs.

BROKEN OPEN 15 September, after sitting untouched because Momple was a name with nothing attached. One NAAIRS query attached a family. TAB WLD 3046/1959: DIVORCE, LOUIS JAMES MOMPLE ROBERTSON VERSUS FLORENCE PERTHINIA — A JOHANNESBURG MAN CARRYING MOMPLE AS A MIDDLE NAME AND ROBERTSON AS A SURNAME, which is the classic signature of a mother's maiden name carried forward and exactly the joint this archive needed, since Phyllis Agnes ROBERTSON's mother was a MOMPLE. He is very probably her brother or first cousin. And behind him a whole family with estates: MHG 6060/50 MOMPLE, LOUIS ARTHUR, surviving spouse DORA MOMPLE BORN CARLSE; MHG 695/64 Dora Momple; TPD 1468/1950 Douglas Arthur Momple in re the late Louis Arthur Momple; and three daughters married out as Archer, Ackhurst and Gervais. MHG 6060/50 IS INSIDE THE ESTATE INDEX'S RANGE AND CAN BE ORDERED — its death notice names Louis Arthur Momple's parents and its estate file lists his children. If Phyllis's mother is among them, row 52 is answered and the Mauritian line behind her opens with it. || REASSIGNED 16 September. Everything reachable from a desk has been done: the family is identified, the surname joint is found, and the estates are located. WHAT IS LEFT IS AN ORDER, NOT A SEARCH — MHG 6060/50, Louis Arthur Momple, died 1950, surviving spouse Dora born Carlse. It is INSIDE the NAAIRS window and its death notice names his parents while the file lists his children. It should be added to REQUEST 20's Pretoria bundle. The one desk route left, the 1925 and 1931 Johannesburg voters rolls identified in row 89, sits behind a GenDatabase membership.

WORKED, and the row turned out to be WRONG. It said to send a short amendment rather than let it sit. Reading the thread first showed the researcher already has two emails from 15 September — the Sunday brief at 23:15 and an addition at 03:25 that ALREADY asks the Deeds Registry question and names the 1948 and 1956 contracts — and has replied to neither, and has not said whether she takes the work at all. Our own second letter says in its first paragraph that everything in it can wait until she has answered about rates. A third email the same day would contradict that sentence, and would cost nothing to avoid, because she cannot go to the Deeds Registry before terms are agreed. DRAFTED AS REQUEST 18 AND HELD, to be appended to the reply when she answers rather than sent on its own. This is exactly what row 49 is for.

STRUCK within the hour, and it is worth saying why. The Find a Grave record's PLACE field offers five candidate cemeteries for the ambiguous string 'West Park Cemetery' — Johannesburg, Cleveland Ohio, Long Eaton Derbyshire, Hereford Texas, and Bulawayo — and FamilySearch surfaced the Johannesburg one only because the query was filtered to South Africa. The record's own Country, State, County and City fields resolve to UNITED STATES OF AMERICA / TEXAS / DEAF SMITH / HEREFORD. He also appears as a father in Illinois births of 1883. He is an American and has nothing to do with this family. A place-name collision in an index, caught before anything was published — which is the first time this week that has happened before rather than after.

DONE 15 September — overtaken rather than answered, and the detour was worth more than the row. The gazette gives nothing for 64 Greenfield Road, so instead of hunting the address the CEMETERY REGISTERS were opened, on the reasoning that a burial register records where the dead lived and a grave register records who BOUGHT the grave. Both are true. West Park Cemetery Private Graves Register, folio W 8 (published here first as W58; corrected the same evening): purchaser WEAR, Mr L. A. R., of 271 WILSON STREET, FAIRLANDS, receipt 11484, 13 September 1969, sixty pounds, grave 56045 section E.C., person buried HENRY ROBERT WEAR. Read at five times magnification the initial is a cursive L, not a J. And the 1986 burial register gives Martha Maud, aged 90, died at the Hillcrest Geriatric Centre, usual residence 12 and a half Second Street, PARKHURST — buried in grave 56045, the same grave. Row 55's own question is still unanswered; it has been overtaken, because the archive now has an address for LEONARD, which is what it actually wanted.

A reply came in at 06:54 on 15 September and was sitting unread. Maria Joao Pires, Coordinator of the Arquivo da Defesa Nacional, writes that for military records one must contact the ARQUIVO GERAL DO EXERCITO, arqgex@exercito.pt — a different body from the Arquivo Historico Militar this archive had already written to. Written to the same hour, citing the referral by name and date, Gmail 1a0a40447162eada. The prize is the same as it always was: a Portuguese military recenseamento names the father AND the mother of the man recensed, and Henrique's father is unknown. Row 33 makes this more important, not less — the South African civil registers have no parentage field at all, so the Portuguese paperwork is the only route left for him. Acknowledged the same day at 07:42: "Confirma-se a receção da sua correspondência eletrónica a qual será tratada pelos nossos serviços." So the referral address works. An acknowledgement, not an answer.

DONE 15 September, and the answer is that THE RECORD WAS NEVER MADE. Four more West Park images were rebuilt tile by tile. (1) The full two-page spread of Martha Maud's 1986 burial, folio W 104, image 3:1:3QHV-B3PY-2FKN, 7,407 x 3,164, 377 tiles: cause NATURAL CAUSES, burial order 37640, 17.3.86, undertaker REG, receipt 61610, R70.00, nature of interment "New Reopen." — and the Remarks column, which is where an authority would be noted, is EMPTY for her row. Not illegible, not cropped: empty. (2) The index side of the Private Graves folio, image 3:1:3QHV-13P1-89VT, read line by line across three magnified passes, holds exactly ONE Wear: Henry Robert, 13.9.69, grave 5604/5, Fol. W56. MARTHA MAUD IS NOT IN IT. A 1986 interment into a 1969 private grave generated no new private-grave entry, which is itself the explanation for the blank Remarks column: the reopening was handled as a burial, not as a grave transaction, so nobody had to sign for it. The row closes with the reason attached, not struck. What is left is the undertaker's own file — the register names REG — or Martha Maud's estate at the Master of the High Court, and neither is worth a row until the Master answers what it already has. THE DETOUR PAID AGAIN, twice: burial register folio W 56 row 4 gives HENRY ROBERT WEAR's usual residence, 5 LIVINGSTONE CT., 4TH ST., PARKHURST — the first address ever recorded for him in South Africa — with address at time of death General Hospital Jhb., died 11.9.69 of BRONCHOPNEUMONIA aged 80, burial order 2894, undertaker Cooks, R15.00. And the folio number this archive published for the grave register was WRONG: it is W 8, not W58, confirmed three ways.

DONE 15 September, and it paid far beyond the address. One gazette query on "Livingstone Court" returned four South African hits; the 1965 and 1977 ones are other people, but GOVERNMENT GAZETTE 4620 OF 21 MARCH 1975, under NATURALIZATION NOTICES, carries: EVANS, Mrs Janette Eileen Maud, formerly Bailey, nee WEAR, saleslady, 5 LIVINGSTONE COURT, 28 VIERDE STRAAT/FOURTH STREET, PARKHURST, JOHANNESBURG. The OCR interleaves two columns and the entry had to be unpicked from its neighbour, so the page itself was read. And the grant followed in GAZETTE 5007 OF 12 MARCH 1976: R54033 - EVANS, Janette Eileen Maud, 28/10/19, England, Citizen of the U.K. and C., 30/10/75. THREE GAINS. (1) The Parkhurst address is corroborated from a wholly independent source six years later, with the building's street number added — the cemetery clerk's hand and the Government Printer's type agree. (2) The birth date 28 October 1919 falls inside GRO quarter Q4 1919, so the Hitchin birth and the Johannesburg naturalisation are the same woman without an assumption. (3) THE FAMILY STAYED BRITISH: Leonard's eldest sister was still a citizen of the United Kingdom and Colonies in 1975, forty years after the household left Hertfordshire. That explains why a man in Johannesburg from the late 1930s to 1969 leaves so little in South African government series. Also new: she was married twice, formerly BAILEY. Checked and rejected along the way: "Wear, Martha" in Gazette 2438 of 20 June 1969 is MARTHA CUMMINGS WEAR, BORN CAMASK, of 17 Botha Avenue, Witbank, estate 6167/66 — not Martha Maud born Castle. Opened rather than trusted, which is now the rule.

NEW, and it follows from the parents being at a DIFFERENT address on the same day. On 13 September 1969 Henry Robert Wear's household was a Parkhurst flat and L. A. R. Wear's was 271 Wilson Street, Fairlands. So the Fairlands address is Leonard's own, not a borrowed family one — which makes it worth testing in a way a family address would not be. A Deeds Office search on the erf will say whether he held title, and if he did it will say when he bought and WHEN HE SOLD OR DIED OUT OF IT, which bears directly on 1973 against 1979. Costs a search fee and needs somebody in Johannesburg; add to the held researcher brief rather than opening a correspondence. || REASSIGNED 16 September: this was never a desk job and should not have sat under "this archive". A Deeds Office search on the erf needs a person in Johannesburg. It belongs with REQUEST 14's Deeds Registry question, which already asks about the ante-nuptial contracts, rather than standing alone.

NEW and deliberately small. Row 4 of burial folio W 56 carries two lines in its Remarks column: "P. Fol. W.8.", which is the private-graves cross-reference and is certainly Wear's, and above it, in a smaller darker hand, "Transf. P.Fol. F 3/8." — a transfer of private rights to or from folio F 3/8. The ruling below both lines puts them in row 4's cell, but the page curves across two thousand pixels and the upper line may belong to row 3. THE ARCHIVE IS NOT CLAIMING IT. If it does belong to Wear then the grave rights in 56045 were transferred, and whoever they were transferred to is a name this archive does not have — which is the question row 57 asked and could not answer from the books it opened. Resolving it means finding the F-series folio in the same film and reading what grave it covers. One image, no fee. 15 September, partly answered without opening the F folio: the Private Graves Register HAS an "Authority to Purchase — Next of Kin" column and IT IS BLANK, for Wear and for six consecutive neighbouring rows (receipts 10997, 10999, 11009, 11475, 11484, 11517). The clerk on folio W 8 never used that column. So even had the 1986 reopening been recorded, the register's relationship field was not being filled in. The F 3/8 folio is still worth one image, but it is now the only thing this book can still give. || BLOCKED 16 September: THE FAMILYSEARCH SESSION HAS SIGNED OUT and this row cannot move without it. The catalogue and persona APIs now return 401 and the site redirects to the login page. This archive will not enter credentials on David's behalf. One sign-in in the browser unblocks this row, rows 42, 60 and 88 together, and the post-1970 marriage register test in row 81. UNBLOCKED 17 SEPTEMBER: the sign-in is live, and the row now has its image references. The West Park pages are ark 3:1:3QHK-Q3PY-2FQG and 3:1:3QHV-13PY-2FCC for Henry Robert in 1969, and 3:1:3QHV-B3PY-2FKN for Martha Maud in 1986, all in Johannesburg Cemetery Records, collection 4453927. Written into data/familysearch-sa-family.tsv so the next attempt starts inside the film rather than looking for it. What still has to be done by eye is finding the F-series folio in the same film and reading what grave it covers. NOT DONE TODAY, and said plainly rather than left looking done. || BLOCKED AGAIN 20 SEPTEMBER: THE FAMILYSEARCH SESSION HAS SIGNED OUT for the second time in four days. The search page now says 'Create a free account or sign in to see all available results' and the persona API returns 0 results for a query that returned 111 on 17 September — a zero that means signed out, not absent, which is exactly the kind of false negative this archive keeps catching. I do not sign anybody in. One sign-in in the browser unblocks this row and row 174 together. || 20 SEPTEMBER: THE SIGN-IN IS BACK AND THIS ROW WAS STILL NOT DONE, which is said plainly rather than left implied. The session was spent on the Ermland parish registers instead, because Lizzy's birthplace surfaced the same morning and was the larger prize. The image references captured on 17 September stand and the row is ready: the F-series folio has to be found by eye in the same film as ark 3:1:3QHK-Q3PY-2FQG. NOT A NEW BLOCK — A CHOICE, AND THE ROW SHOULD NOT PRETEND OTHERWISE. AND THEN IT WAS ATTEMPTED AFTER ALL, AND GOT MOST OF THE WAY. The row has never named the film it needs. IT NOW DOES: the catalogue record behind the indexed collection is olib:4148464, 'South Africa, Johannesburg City Cemetery records', TWO HUNDRED AND TEN FILMS, twelve of them West Park private graves registers by year range. LEONARD BOUGHT GRAVE 56045 ON 13 SEPTEMBER 1969, SO FOLIO W 8 IS IN FILM 107638608, 'WPC PRIVATE GRAVES REGISTERS 1963-1970', 488 IMAGES — and nowhere else. The film was opened and its structure read off the page: IT IS ALPHABETICAL BY PURCHASER SURNAME, image 401 is the T section (Theunissen, Tyson, Terblanche, Tennyson, Taljaard, Tomlinson, Taylor), the printed form is 14/6082 and the columns are NAME OF PURCHASER and ADDRESS. So THE F FOLIOS SHOULD SIT AROUND IMAGES 180 TO 200 and the W folios around 440 to 470. FOLIO F 3/8 WAS NOT READ: the image viewer kept reverting to its thumbnail grid, each page load costs fifteen to twenty seconds, and the hunt was abandoned rather than padded out. THAT IS A FAIR TRADE AND THE ROW SAYS SO: the transfer line may not be Wear's at all — the ruling puts it in row 4's cell but the page curves and it may belong to row 3 — and a film number plus an alphabetical structure is worth more than another hour of clicking, because it makes every future West Park question cheap. In data/westpark-films.tsv.

NEW, and it is a record class this archive has never used. A gazette naturalisation notice is a name and an address; the FILE behind it records date and place of birth, previous nationality and PARENTS. Row 33 established that the South African civil marriage registers have no parentage field at all — this series does. Certificate R54033 was granted to Janette Eileen Maud Evans, nee Wear, on 30 October 1975; if the file names Henry Robert Wear and Martha Maud Castle it is the first South African government record to state that couple as the parents of a child. It does not prove Leonard was theirs. THE SECOND QUESTION IS THE BIGGER ONE: ask in the same letter whether any naturalisation file exists for WEAR, Leonard Aubrey Robert, born 1917. A phrase search of the gazette returns nothing for him and no Wear in either the 1975 notice list or the 1976 grant list, but the lists are alphabetical and the OCR interleaves columns badly enough that this very entry had to be unpicked by hand — SO THAT NULL IS NOT AN ANSWER. Files for this period sit with the Department of Home Affairs, older series transferred to the National Archives, Pretoria. A letter, not a purchase. Send with the Master of the High Court business. || WORKED 16 September to the point of deciding WHO to ask, and the answer is nobody new. NAAIRS does not index individual naturalisation files — its 342 NATURALIZATION documents are policy files and bills — so the file behind certificate R54033 sits with the Department of Home Affairs or in an unindexed National Archives series. THE ARCHIVE ALREADY HAS TWO LIVE THREADS THAT COULD CARRY THE QUESTION: Enquiries2@dsac.gov.za, which answered within two days on 12-14 September, and Ms Mavis Xaba at mavisx@dsac.gov.za, WHO HAS ALREADY BEEN WRITTEN TO TWICE AND MUST NOT GET A THIRD COLD THREAD. So this is not a new letter: it is one added question on the existing NARS thread, asking whether naturalisation files of the 1970s are held by them or by Home Affairs, and whether the file behind a known certificate number can be produced. Cheap, and it avoids the duplication this archive has already been burned by once.

DONE 16 September, AND IT IS THE BEST THING FOUND THIS WEEK. The archive did not know which country the Bailey marriage was in, let alone which year. It is the FIRST of the thirty-six, and it is ENGLAND. FreeBMD, Marriages, December quarter 1939, >99% complete: WEAR, Jeanette E M — spouse BAILEY — HITCHIN — vol 3a page 4416; and the reverse, BAILEY, EDWARD J — spouse WEAR — same volume and page. HITCHIN IS THE FAMILY'S OWN REGISTRATION DISTRICT, where all six Wear children were born. TWO CONSEQUENCES. First, it rewrites the 1939 Register. That register, taken 29 SEPTEMBER 1939, found her alone at 18 Ridge Avenue, Letchworth, and this archive read it as "and one daughter stayed" — the household gone to South Africa, a girl of twenty left behind. SHE WAS NOT LEFT BEHIND; SHE WAS STAYING TO BE MARRIED, within three months and probably less. Second, and this is the point: AN ENGLISH MARRIAGE CERTIFICATE NAMES BOTH FATHERS. Row 33 established that the South African register has no parents' column at all — which is why the whole sibship rests on a seven-person family tree, why request 19 goes to the Master for three estates, and why row 61 chases a naturalisation file. This certificate costs about £14 and comes by email.

NEW, and it is a fifth court file and a year the archive did not have. TAB, WLD 978/1960: ILLIQUID CASE, PAYMENT, MAWTHANE DABEAH KWELE VERSUS L A R WEAR. Leonard was in Johannesburg in 1960 and somebody sued him for money. The plaintiff's name is African, and a Black plaintiff bringing a payment action against a white defendant in the Witwatersrand Local Division in 1960 implies a working or trading relationship rather than a casual one — which would make this the only surviving trace of what Leonard actually DID for a living between the 1951 register's ENGINEER and the 1969 grave. The archive is not guessing further; the file would say. Add to the held researcher brief alongside the ante-nuptial contracts, since it is the same reading room. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20.

NEW. WLD 2960/1962 is Marjorie Eileen Wear divorcing Leonard; WLD 2961/1962, filed one number later, is the same plaintiff suing ANOTHER WOMAN for DAMAGES. Same plaintiff, consecutive numbers, same year: in the practice of the period that is an action against the co-respondent, and it means the second marriage ended over somebody the archive has never heard of. HER NAME IS IN THE INDEX AND DELIBERATELY NOT ON THE SITE — she would be in her eighties or nineties if living, there is no death record for her, and the claim is an imputation of adultery drawn from two case numbers rather than from the file. Held in the private notes. What would settle it is the file itself, and it matters beyond gossip: a woman Leonard was involved with in 1962 is the likeliest person to have been near him in 1969, and 1969 is where he disappears. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20.

DONE 15 September, and the answer is a bounded negative that is itself a fact about her. The gazette was swept YEAR BY YEAR 1946-1980 for the forename Ivone — the search API takes a year filter and pages, which is how a common forename becomes tractable. Ten South African gazettes carry it in that window and EVERY ONE WAS OPENED: a Radium Station general dealer in 1951, Ivone da Conceicao Isaac in the 1962 grant list, Maria Ivone Freitas Basilio de Abreu of Emmarentia in 1967, Ivone Bertholet de Oliveira e Sousa Duran of Lourenco Marques in 1968, Maria Ivone Silva Almeida in the 1968 grant list, Maria Ivone Melo Gouveia in the 1970 list. NONE IS HER. The exact forms "Salkow, Ivone", "Ivone Salkow", "LUWINSKI, Ivone", "MOTA FIGUEIREDO" and "Rocha, Ivone" return zero each. IVONE DID NOT NATURALISE AS A SOUTH AFRICAN under any of her three surnames. The sweep also taught the format — surname, forenames, date of birth, country of birth, previous nationality, certificate number — and showed the lists FULL of Portuguese and Madeiran names through exactly the years she lived there. Other Portuguese naturalised steadily; she did not. That fits a woman who went back to Portugal and died at Porto, and it makes the Portuguese correspondence more important, not less.

DONE 15 September. "da Rocha, Henrique" and "Henrique Figueiredo Pinto" return three South African gazettes and ALL THREE ARE THE 2023 ESTATE NOTICES THIS ARCHIVE ALREADY HAD for estate 12294/2023. No naturalisation notice, no grant. The route is closed and the father the tree calls Unknown is still unknown. What remains for him is the Portuguese military recenseamento, which is row 56, and the South African estate file itself.

DONE 15 September. "Salkow, Morris", "Morris Isaac Salkow" and "Salkow, Mrs" all return zero. The bare surname returns scores of hits across 1941-1969 but they are trade and business notices, not persons. No naturalisation for him either. Where Morris Isaac Salkow was born is still unestablished, and the gazette will not say.

WORKED 16 September to the point where only a letter remains, and the letter is written. REQUEST 19 IS DRAFTED AND NOT SENT, because this archive's own rule after the Xaba duplication is to verify the address before writing: the Johannesburg Master's office has not been checked the way Cape Town's was for request 11. THE ROW ALSO GAINED A THIRD SISTER. Cynthia Yvonne Elizabeth Halsey, born Wear, b.1931, DIED 2000 — the death was not previously fixed. And KATHLEEN EDITH FOOT, born Wear, b.1921, DIED 2010, married James Cecil Foot in 1945, is flagged in the ESTATE FILE index as well as the death index, which makes her the likeliest of the three to have a file that survives. CHECKED SO THE LETTER DOES NOT ASK WHAT THE SENDER COULD FIND: none of the three has a Government Gazette estate notice — "CYNTHIA YVONNE ELIZABETH" returns nothing at all, "JANETTE EILEEN MAUD" returns only her 1975 naturalisation, "KATHLEEN EDITH" with FOOT returns nothing in 2009-2014. That is expected rather than suspicious: a J193 advertisement is required only above a value threshold. So they must be asked for by name and date of death, which is what request 19 does. || ADDRESS CHECKED 16 September, and the check changed the letter. MastersJohannesburg@justice.gov.za is NOT IN USE — the department's own PDF says so in capitals. The live contact is MorSello@justice.gov.za, Ms Morwanki Sello, controller for file numbers ending in 4, AND THIS MAILBOX ALREADY WROTE TO HER on 15 September about estate 12294/2023. So request 19 must go in as a FOLLOW-UP ON THAT THREAD, not as a cold letter, or it repeats the Xaba duplication that caused the shared-contacts rule to exist. And a structural problem the letter now owns openly: the office allocates by the last digit of the estate number, and this archive HAS NO ESTATE NUMBERS FOR THE THREE SISTERS — that is what it is asking for — so Ms Sello may have to redirect it, and the letter says so rather than pretending the choice of desk was informed.

FOLDED INTO REQUEST 19 with row 68 and its two sisters — one letter, three estates, because they are three sisters at one Master's office and three separate approaches would read as duplication. Janette Eileen Maud Evans, formerly Bailey, born Wear, 28 October 1919 to 2001. Her birth date is exact because her own naturalisation certificate R54033 of 30 October 1975 prints it. Her estate is NOT gazetted, so it must be asked for by name. If the Master declines, the fallback is row 61 — the naturalisation FILE behind R54033, which records parents and sits with Home Affairs or the National Archives, not with the Master. || ADDRESS CHECKED 16 September, and the check changed the letter. MastersJohannesburg@justice.gov.za is NOT IN USE — the department's own PDF says so in capitals. The live contact is MorSello@justice.gov.za, Ms Morwanki Sello, controller for file numbers ending in 4, AND THIS MAILBOX ALREADY WROTE TO HER on 15 September about estate 12294/2023. So request 19 must go in as a FOLLOW-UP ON THAT THREAD, not as a cold letter, or it repeats the Xaba duplication that caused the shared-contacts rule to exist. And a structural problem the letter now owns openly: the office allocates by the last digit of the estate number, and this archive HAS NO ESTATE NUMBERS FOR THE THREE SISTERS — that is what it is asking for — so Ms Sello may have to redirect it, and the letter says so rather than pretending the choice of desk was informed.

Row 36 established that the NAAIRS deceased-estate index STOPS AT THE END OF 1976, so the absence of a 1986 estate proves nothing: it is out of range, not missing. Martha Maud Wear died 16 March 1986 aged 90, usual residence 12½ Second Street, Parkhurst, and was buried in her husband's grave. Her estate would be at the Master of the High Court, Johannesburg. Its death notice names HER parents — the Castle side — and its next-of-kin and heirs schedule lists her surviving children, WHICH IS THE LIST THIS ARCHIVE MOST WANTS. If Leonard predeceased her the schedule normally says so and often gives the date. Fold into the Master correspondence already running as row 2.

WORKED 15 September to the end of what is free, and the result is a second instrument on which 1973 fails. The England and Wales probate service is free and its images ARE viewable without an account — but IT DOES NOT NAME-INDEX ANYTHING BEFORE 1996. A surname-and-year search returns scanned calendar pages and matches the string anywhere on the page, so WEAR also catches Weardale and Wearmouth. Searched with the keyword AUBREY to cut the noise: 1973 returns ZERO PAGES; then 8, 24, 19, 31, 13, 23, 19, 13 for 1974 to 1981. THE FAMILY TREE'S YEAR RETURNS NOTHING, on a second independent index, the evening after the South African estate index said the same. The 1979 pages have NOT been read — twenty-three dense calendar pages is an afternoon by eye, or minutes on a commercial index that has keyed the calendar by person. The row stays open and now says what it would cost. || CLOSED 16 September by row 73 rather than by reading the images. If he did not die in England or Wales between 1970 and 1985 — and a 99%-complete index that prints dates of birth says he did not — then there is no English death for an English grant of probate to follow. The 23 unread pages for 1979 would be a search for something now known not to be there. ONE CAVEAT KEPT, because it is real: an English grant CAN follow a foreign death, and the calendar records those in a recognisable form — a page read on the way records ADAMSON, Charles Mary, "of 126 The Gardens Esplanade Durban SOUTH AFRICA died 7 April 1979 administration with will Pietermaritzburg £3500 in England and Wales SEALED London 3 December 1981". So a resealed South African grant would appear. That is a narrower search than the one this row began with, and it is the only reason not to mark the row struck.

STRUCK 16 September — not abandoned, but CLOSED STRUCTURALLY, and it is worth more struck than it was open. Row 73 promoted this to the most important untested thing about him. It turns out the series CANNOT CONTAIN HIM, for three separate reasons carrying three different weights. (1) FIRM: the GRO consular death registers run 1849 TO 1965, and both candidate years are outside them. (2) FIRM, from gov.uk itself: the modern statutory service covers deaths "on or after 1 JANUARY 1983" and is optional — 1973 and 1979 are both before it. (3) PROBABLE BUT NOT PROVEN, and marked as such because it is a council web page and not legislation: Coventry City Council's guidance lists SOUTH AFRICA among countries where a death CANNOT be registered with the British authorities at all, alongside Australia, Canada, Bermuda and Zimbabwe — which is what one would expect of Commonwealth countries with their own civil registration. ANY ONE OF THE THREE WOULD BE ENOUGH; ALL THREE AGREE.

DONE 16 September, and it is THE BEST-BOUNDED NEGATIVE THIS ARCHIVE HAS EVER PRODUCED ON HIM. Row 58's finding that the family stayed British raised a hypothesis nobody had entertained — that Leonard went home to die. HE DID NOT. FreeBMD transcribes the GRO index; its coverage chart for deaths shows EVERY QUARTER FROM 1970 TO 1983 AT 100%, and the search prints the figure beside each quarter: 58 CONSECUTIVE QUARTERS, EVERY ONE MARKED >99%. And from 1969 the GRO death index carries FULL FORENAMES AND THE DATE OF BIRTH, so he can hide behind neither an initial nor a common surname. Searched WEAR, deaths, March 1970 to December 1985, no forename: 241 ENTRIES, SIXTEEN OF THEM BORN 1915-1919, AND NOT ONE IS HIM. The only 1917 birth is a Sam Wear of Halifax. A separate search on the forename Leonard returns exactly one man, Leonard Rigby Wear born 10 October 1897, died Claro June 1970 — twenty years too old, wrong county. SETTLED: he did not die in England or Wales between 1970 and 1985, in either candidate year or any year between.

DONE 16 September, and it paid for itself on the fourth surname. THE FIND IS WLD 2225/1953: ILLIQUID CASE. DIVORCE. MORRIS ISAAC SALKOW VERSUS IVONNE (BORN DE FIGUERIDO). THEY MARRIED ON 13 FEBRUARY 1953 AND HE SUED HER FOR DIVORCE IN THE SAME YEAR. The archive knew the marriage ended — FamilySearch has him marrying in 1943, 1953 and 1960 — but not that it ended almost at once; Johannesburg case numbers ran to five figures by the 1970s, so 2225 of 1953 is not a December filing. Three consequences: it removes a stepfather the archive had been assuming was a constant through Derrick's childhood; it explains the hole after 1953 as a fact about her life rather than a failure of searching; and it is a new orderable document naming Ivone, whose summons would recite the date and place of the marriage, both ages and domiciles, and the children by name. ALSO SWEPT: NORQUOY gave Robert Selwyn (row 82); MOMPLE gave the Robertson joint (row 39); CASTLE gave the Longhurst coincidence; SALKOW gave nine Johannesburg estates all inside the index window, one of which is probably Morris Isaac's parent; and LEIBHOLZ returns ZERO across all repositories, which is expected — they were never in South Africa — and is recorded so nobody runs it again.

DONE 15 September. All six rows of litigation.json verified directly against NAAIRS rather than against a commercial mirror or a photograph of somebody's screen. TPD 534/1948 no longer needs its excuse: the index gives it in full — TAB, TPD, VOLUME 5/795, ILLIQUID CASE. DIVORCE. JOHANNES ANTONIE OELOFSE DE WITT VERSUS YVONNE (BORN DE FIGUEIREDO) — and THE VOLUME NUMBER IS NEW, which is what an order form actually needs. The nourl excuse is removed. And WLD 2545/1950 explains its own earlier obscurity: the index spells her IVONE LUWENSKI, not Luwinski, so a correct-spelling surname search could never have found it.

DONE 15 September, and the answer is that KURT'S ARRIVAL IS NOT IN THE INDEX AT ALL. The PIO immigration series is real and indexed at item level, and it holds exactly one Wear and no Luwinski. KAB PIO 907, reference 38964E, IMMIGRATION PAPERS H C WEAR, 1949 — almost certainly not ours: the initials are H C, not H R or L A R, and 1949 is a decade after this archive dated the family's departure. Recorded because it is the only one and because "almost certainly not" should be written down rather than decided twice. Also found: KAB PIO 82445E, immigration papers re EISENHARDT, 1957. IMMIGRATION+LUWINSKI and PIO+LUWINSKI both return nothing. Kurt has no ship, no permit and no date, and the index that would hold them does not have him.

DONE 15 September, and the answer is that LEONARD IS NOT THERE. Searched by collection rather than by guesswork — f.collectionId=4453927, the Johannesburg cemetery registers, 517 raw hits on the surname, collapsed to six distinct people: HENRIETTA WEAR buried 1909, HENRY ROBERT WEAR 13.9.1969, MARTHA MAUD WEAR 19.3.1986, EDWINA ANN WEAR 8.7.2014, MICHELLE ANN WEAR born 1978 died 27.3.2017, and DAN B WEAR undated. He bought grave 56045 and he was not buried in it, and he is in no other Johannesburg cemetery register the collection indexes. BOUNDS: this is the INDEXED layer of ONE collection covering 1840-2019; it does not cover unindexed images in it, other Johannesburg cemeteries, or the Chevrah Kadisha, which is a separate database — a Wear would have no business in a Jewish cemetery but the exclusion is stated rather than assumed. TWO WEARS ARE NEW: Michelle Ann Wear, born 1978, the right age to be a granddaughter of Bruce Wallace Wear who married four times; and Dan B Wear. Both written down, neither pursued.

Two people are in that grave and a third person paid sixty pounds for it. A private grave of 1969 in section E.C. would ordinarily carry a memorial, and the burial register has a Remarks column headed "Private Rights, MEMORIALS" in which other rows on the same page record memorials erected and their cost. A photograph of the stone would say who is commemorated and, if a family stone, possibly who paid for it and when it was last added to. Ask the City of Johannesburg cemeteries office, or any of the local volunteer photographers who work West Park. Costs nothing but a request.

WORKED 16 September to the limit of what can be established from here, and it is LESS TRACTABLE THAN IT LOOKED. "COOKS" is a specific firm named against Henry Robert Wear in September 1969 and nothing else on that page. "REG", which the 1986 register gives for Martha Maud, IS ALMOST CERTAINLY NOT A FIRM AT ALL — it appears against most rows on the 1969 page as well, which is not how an undertaker's name behaves. It is far likelier to be a clerk's abbreviation, and the archive WILL NOT GUESS WHICH: reading it as a company and writing to that company would be the Caroline Walls error again. Doves, founded 1882 and headquartered in Johannesburg, is the obvious large survivor and could be asked, but there is no evidence connecting it to either entry. WHAT THIS ROW ACTUALLY NEEDS IS THE REGISTER'S OWN KEY — the abbreviation table a cemetery clerk would have used — or one question to the City of Johannesburg cemeteries office, which is the same office row 78 must approach for the photograph of grave 56045. FOLD THE TWO REQUESTS INTO ONE.

The archive now knows of five and has read none: WLD 2026/1951 (Phyllis v Leonard, divorce), WLD 790/1957 (Leonard v Norquoy, opposed application), WLD 978/1960 (Kwele v L A R Wear, payment), WLD 2960/1962 (Marjorie v Leonard, divorce) and WLD 2961/1962 (Marjorie v a named woman, damages). A divorce summons of the period recites BOTH PARTIES' FULL NAMES, AGES, DOMICILES AND THE DATE AND PLACE OF THE MARRIAGE, and often the children by name and birth date. Five files from one depot on one visit; the researcher is already holding a brief for the ante-nuptial contracts and this belongs in the same envelope. Supersedes the separate handling of rows 63 and 64. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20.

ATTEMPTED 15 September and BLOCKED on image availability, not on reasoning. The test needs a post-1970 South African marriage register image, and the only Wear marriage FamilySearch has indexed with a register image is the WRONG SIDE OF THE LINE: Bruce Wallace Wear and Jillian Anne Thorp, 1959, ark 3:1:3Q9M-CSVT-C992-Q, which this archive already has. His 1977 divorce from Margaret Gannon and his 1980 marriage to Edwina Ann Green are known from GenDatabase and from NAAIRS, NOT from a filmed register, so there is no image to open. The row stays open and now names what it needs: any South African civil marriage register image dated after 1970 — it does not have to be a Wear. The question is about the FORM, not the family, and one page of any post-1970 register would settle whether row 33's finding needs its date range narrowed in public. STRUCK 17 SEPTEMBER, AND FOR A REASON WORTH MORE THAN THE ANSWER. The row needs any South African civil marriage register image dated after 1970. THERE IS NONE TO REACH BY NAME, because "South Africa, Civil Marriage Records, 1801-1974" IS A NAME INDEX TO 1960, NOT TO 1974. Measured, not assumed: exact-year counts run 133,215 for 1950, 148,400 for 1955, 171,201 for 1959, 178,037 for 1960 - AND THEN 124 FOR 1961, 37 for 1962, 6 for 1966, and nothing at all after 1968. A cliff, not a decline. The 2.69 million images may well run to 1974; the index that lets you find one does not. So this row is not blocked on a sign-in and never was - it is blocked on a boundary inside the collection, and it can only be answered by browsing images or by a register held elsewhere.

ADVANCED 15 September — a candidate name, from NAAIRS. Phyllis remarried a Norquoy on 14 June 1952 and Leonard sued "Phyllis Agnes Norquoy" in 1957. NAAIRS holds fourteen Norquoy documents and two are a Johannesburg man in exactly the right years: WLD 1198/1957, Edward William Allen versus ROBERT SELWYN NORQUOY, and TPD 1624/1960, Sadi Jacobs versus Robert Selwyn Norquoy. ROBERT SELWYN NORQUOY IS VERY PROBABLY HER SECOND HUSBAND — rare Orkney surname, right city, right decade, and WLD 1198/1957 filed the same year as Leonard's WLD 790/1957 against her. Stated as an inference from surname, place and date; no document yet joins them. Next: the 1952 marriage register under the name Norquoy, and his estate. There is no Norquoy estate for a Phyllis Agnes in the index, but the index stops in 1976 and she may have outlived it. || BLOCKED 16 September, and the blocking is specific. She is NOT in the South African death databases under NORQUOY — thirty-one Norquoy records and not a Phyllis among them — so either she died elsewhere, or remarried again, or the index does not have her. The remaining free route is her 1952 marriage register: South African civil marriages are on FamilySearch to 1974, so a June 1952 Johannesburg marriage SHOULD be reachable — but the FamilySearch session has signed out and this archive will not enter credentials. One sign-in unblocks it, together with rows 42, 60, 81 and 88. THE NORQUOY FAMILY IS NOW A SHORT LIST INSTEAD OF A SURNAME, 17 September, from the free gazette archive. All 41 gazette documents carrying the name were read. IN JOHANNESBURG: THOMAS SMITH LEASK NORQUOY, ambulance officer, estate 3529/44, liquidation account gazetted 1945; SOLDELINA O'NEILL NORQUOY, BORN 14 SEPTEMBER 1900, estate 16050/83, of BRAMLEY HOUSE, 71 LINDEN ROAD, BRAMLEY; LOUIS WILLIAM NORQUOY, a retired miner, estate 1638/55; ROBERT SELWYN NORQUOY; WILLIAM KNIGHT NORQUOY; PETER NORQUOY; BARBARA ALICE NORQUOY; and later Shaun Thomas and Michelle. PHYLLIS MARRIED A NORQUOY ON 14 JUNE 1952 AND ONE OF THESE MEN IS PROBABLY HIM. THIS ARCHIVE IS NOT GOING TO GUESS WHICH — that is exactly how the Giddins mistake was made eight days ago. AND SHE IS NOT IN THE GAZETTE HERSELF: Norquoy AND Phyllis returns nothing; Norquoy AND Agnes returns three documents and all three are coincidental page-mates. WHAT WOULD SETTLE IT: the 1952 marriage register, which names the groom in full — it is already on the researcher's list as part of the Wear court files, and WLD 790/1957 names her as Norquoy, so the summons in that file will carry the surname's owner.

DONE 16 September, end to end, and she turns out to be the best-recorded person in Leonard's orbit. MARJORIE EILEEN WALKER, born 1916, died 2003. THREE HUSBANDS AND SIX COURT FILES: WLD 185/1949, divorce, Marjorie Eileen DUNLOP (born Walker) versus Robert John; then 30 October 1951, marries LEONARD at Polana Court, Abel Road, Berea; then WLD 2960/1962 divorcing Leonard and WLD 2961/1962 suing another woman for damages one number later; then WLD 2276/1973 and WLD 7613/1973, Roger William Hannay HUMPHREY against her; then death in 2003 as MARJORIE EILEEN HUMPHREY, BORN WALKER, WITH AN ESTATE FILE. TWO THINGS THIS CHANGES. First, she divorced Dunlop in 1949 and married Leonard in 1951, and Leonard's own divorce is WLD 2026/1951 — two people coming out of two marriages within two years of each other. 1948 sits immediately before her divorce, and the archive's blank nine years for Leonard end exactly where her paper trail begins. Second, and this is the point: SHE WAS MARRIED TO HIM FOR ELEVEN YEARS, THE LONGEST DOCUMENTED RELATIONSHIP OF HIS LIFE, AND HER 2003 ESTATE CARRIES A NEXT-OF-KIN AND HEIRS SCHEDULE THAT LISTS HER CHILDREN.

DONE 15 September, same answer and same reason as row 37. No insolvency or sequestration notice under any exact form of his name, and the loose sweeps are unusable because the surname is an ordinary English word. Recorded as a weak null, explicitly, on the bounded-negatives page.

DONE 15 September, and the shop had a life before Kurt. The 1940 page was unreadable in OCR, so the PDF page was rendered at 200 dpi and read as an image. Government Gazette 2825 of 8 November 1940, page VII: NOTICE OF SALE OF BUSINESS — "The business of PICCADILLY HAT SHOP carried on by PICCADILLY DRESS SALOON (PROPRIETARY), LIMITED, at HANNAH COURT, 81 ELOFF STREET, Johannesburg, has been sold and will be transferred to IVY FUCHS, a public trader, who will carry on the said business under the same style or firm at the same address for her own account and benefit as from the 1st of October, 1940." Attorney Israel Katz, Sauer's Buildings. THREE GAINS: the building is named — HANNAH COURT, which this archive did not have; the shop belonged to a company, PICCADILLY DRESS SALOON (PTY) LTD; and KURT BOUGHT IT FROM A NAMED PREDECESSOR, IVY FUCHS, who had held it for thirteen months. He bought a going concern with a history, not an empty name. Also found: 16 May 1952 no. 4846 lists "Piccadilly Hat Co." in a registration schedule under VCR. 429/8/49, a 1949 registration.

She survived Kurt by seventeen years, was the sequestrating creditor against Phillip Crystal in 1972, traded as Pallou's Dress Centre at 18 Wanderers Street in 1975, and is buried at Westpark, grave V-1470. Her estate is past the NAAIRS cut-off of 1976 and so must be asked for from the Master. Its death notice names HER parents, and its heirs schedule records who was alive in 1985 and what they received — which is the nearest thing to a family list the Luwinski side will ever produce.

DONE 15 September, and the technique transferred but the target did not need it. The catalogue search is /service/search/catalog/v3/search?q.keywords=..., whose replies come back as searchHits[].metadataHit.metadata — not the shape the sister archive's parser expected — and the item call is /service/search/catalog/item/<id>. Applied to Johannesburg it finds olib:4453927, "South Africa, Johannesburg, cemetery registers: COLLECTION RECORD, 1840-2019", and the item record says plainly NOT AVAILABLE ON MICROFILM: it is a digital-image collection, so there is no film_note list of volumes to read and the browse-label problem the technique was invented to solve does not arise here. WHAT REPLACES IT IS BETTER: the collection has an id, and the persona search takes f.collectionId, which turns "every Wear in Johannesburg's cemeteries" into one query. That is how row 77 was answered. NOTE A DISCREPANCY WORTH KNOWING: the item's own summary says the collection holds Olifantsvlei registers 2017-2021 while its title says 1840-2019. The title is right — the 1969 and 1986 West Park entries are in it — so THE SUMMARY IS WRONG, and anyone reading the catalogue description alone would wrongly conclude this collection cannot help them.

Everything this archive has found on FamilySearch has come through the indexed name search or through images it knew to open. FamilySearch now runs full text over images that were never indexed, which is a different instrument: it reads the handwriting on pages nobody catalogued. Run WEAR with LEONARD, LUWINSKI, and MOTA FIGUEIREDO against the South African collections. The cemetery register that broke row 55 open was found by opening images by hand; this is the same act done at scale. || BLOCKED 16 September: THE FAMILYSEARCH SESSION HAS SIGNED OUT and this row cannot move without it. The catalogue and persona APIs now return 401 and the site redirects to the login page. This archive will not enter credentials on David's behalf. One sign-in in the browser unblocks this row, rows 42, 60 and 88 together, and the post-1970 marriage register test in row 81. RUN 17 SEPTEMBER, and the answer is a clean negative with a working control. The sign-in this row was blocked on is live, and FamilySearch's full-text search — which reads the WORDS ON UNINDEXED IMAGES rather than a name index — answers. LUWINSKI: 26 hits, EVERY ONE AMERICAN — Milwaukee and Chicago naturalisation petitions, New York passenger lists, Ohio and Minnesota vital records. NOT ONE SOUTH AFRICAN. LUWINSKY: four, all American. LUWENSKI: three, all American. 'PICCADILLY HAT', the name of Kurt's own shop at 81 Eloff Street: nothing but Lancashire trade directories. AND THE CONTROL SAYS THE NEGATIVE MEANS SOMETHING, which is the lesson of the Oppenheimer business. Searching NOOITGEDACHT — a farm name that exists almost nowhere else — returns page after page of SOUTH AFRICAN VOTERS' ROLLS 1925-1931, PRETORIA DISTRICT 1927, VEREENIGING 1931 and TRANSVAAL MILITARY RECORDS 1900-1903. SO THE CORPUS REALLY DOES CONTAIN SOUTH AFRICA, and the instrument really was pointed at it. BUT NOW THE LIMIT, STATED PLAINLY: WHAT IT CONTAINS IS 1925-1931 AND 1900-1903. KURT ARRIVED IN 1933. LEONARD SAILED IN 1937. THE WEARS CAME IN 1939. EVERY PERSON THIS ARCHIVE WANTS ARRIVED AFTER THE LAST ROLL WAS PRINTED. The search is exhaustive over the wrong years, which is not the same as exhaustive. IT IS STILL WORTH RE-RUNNING whenever FamilySearch adds South African material, and the four search terms are written down here so that it takes two minutes next time. || ALSO 17 SEPTEMBER: the ordinary index, which is the other half of this row, was worked to completion in row 161.

DONE 16 September, and the answer REDIRECTS THE INSTRUMENT rather than closing it. Row 41 asserted that directories and voters rolls "all need somebody physically in Johannesburg or Pretoria". Checked: FOR THE WEARS THAT IS RIGHT, AND FOR THE OTHER FAMILIES IT IS WRONG. Street directories — the only South African directory digitised and searchable at ancestors.co.za is PIETERMARITZBURG 1897. There is no Johannesburg directory for the 1940s or 1950s reachable without a reading room, so row 41's nine-year hole stays shut from this direction. BUT THE VOTERS ROLLS ARE ONLINE FOR 1925 AND 1931, indexed by town across 148 places including Johannesburg, Berea, Fordsburg, Jeppes, Denver, Langlaagte and Bezuidenhout, through GenDatabase (membership) and eGGSA. 1925 and 1931 are far too early for the Wears, who arrived in the late 1930s — WHICH IS WHY NOBODY HERE HAS USED THEM. But they are exactly right for the families this week has just opened up: the MOMPLES and ROBERTSONS of row 52, who were Johannesburg people by 1950 and whose parents would be on a 1925 or 1931 roll; and the LONGHURSTS of row 91, whose divorces are 1915 and 1919. A roll gives name, address and occupation by street. THAT IS THE CHEAPEST TEST AVAILABLE for whether Maud Mabel Castle's family were in Johannesburg before the Wears came.

DONE 15 September — /negatives/ is live and in the navigation. Nine bounded negatives, each with three columns: WHAT WAS SEARCHED, WHAT CAME BACK, AND WHAT WOULD STILL FALL OUTSIDE IT. The last column is the point. It holds the distinction between "this is not there", which is a finding, and "I did not find it", which is a shrug. The page also records the two nulls this archive produced by its own error in one evening — a grid read before its async postback had landed, and a form field posted with an empty value — both of which looked exactly like an empty holding and neither of which was. The rule it enforces: do not publish a null you did not bound.

NEW, out of row 39, and deliberately framed as a test rather than a claim. A Maud born Castle was divorcing an Aubrey Leonard Longhurst in Johannesburg in 1915 and again in 1919, two years before Martha Maud Castle bore a son she christened Leonard Aubrey. Martha Maud cannot be that woman — she was nineteen and in Hertfordshire — but a relative could be, and if she was, the archive gains the thing it has never had: A REASON WHY THIS FAMILY CROSSED TO THAT PARTICULAR CITY. Emigrating families go where somebody already is, and the Wear emigration has always been a bare fact with no pull behind it. Three cheap tests, in order: (1) GRO and censuses for a MAUD MABEL CASTLE born in the 1880s in Hertfordshire or Bedfordshire, and whether she sits in the same household or parish as Martha Maud's family at Baldock; (2) MHG 3632/50, Aubrey Leonard Longhurst's estate of 1950, which is INSIDE the estate index's range and orderable, and whose death notice names his parents; (3) WLD 155/1915, whose summons would recite the date and place of the first marriage. If all three come back negative, say so on the bounded-negatives page and close it. || WORKED 16 September to the end of what is free, and test (1) comes back empty. FamilySearch holds NO record anywhere for a MAUD MABEL CASTLE — not in England, not in South Africa, under the full forename pair. A NAAIRS sweep of LONGHURST returns 136 documents and the family is overwhelmingly Cape-based and older — Edwin Richard Percy, Albert, Charles Spurgeon, a customs officer at Port Elizabeth — with nothing joining any of them to a Castle. THAT IS A WEAK NULL and is recorded as one: FamilySearch fuzzy-matches forenames badly and its England coverage is not complete. WHAT IS LEFT IS THE TWO ORDERABLE DOCUMENTS — MHG 3632/50, Aubrey Leonard Longhurst's estate of 1950, inside the index's range; and WLD 155/1915, whose summons would recite the date and place of the marriage. Both need the Pretoria reading room, so this row now belongs with rows 24 and 80 in one visit. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20. It is TIER 3 in that bundle and explicitly the first thing to cut, because it tests a coincidence and not a lead. 17 September: TESTED AGAINST THE ONE SOURCE THAT COULD HAVE PUT A CASTLE IN JOHANNESBURG BEFORE THE FAMILY CAME, and she is not in it. FamilySearch full-text now reaches the SOUTH AFRICAN VOTERS' ROLLS OF 1925-1931. 'Maud Mabel Castle' returns one hit and it is a Washington State birth register; 'Mabel Castle' returns thirty-odd and not one is African. A woman on a voters' roll is an adult British subject resident in the Union, which is exactly what a Castle already in Johannesburg would have been. SHE IS NOT THERE. It does not close the row — she may have arrived after 1931, she may not have registered, and the rolls FamilySearch holds are not all the rolls there were — but the cheapest test has been run and it came back no.

NEW, out of row 83, AND IT IS A SECOND LINE ON THE QUESTION ROW 31 HAS BEEN STUCK ON. Marjorie Eileen Humphrey, born Walker, died 2003 with an estate file. An estate's next-of-kin and heirs schedule lists the deceased's children. She was married to Leonard from October 1951 to 1962 — eleven years, the longest documented relationship of his life. IF SHE HAD A SON BY HIM, THAT SON CARRIES LEONARD'S OWN Y CHROMOSOME. Row 31's candidate is Bruce Wallace Wear, Leonard's brother, who if living is NINETY; a son of Leonard born in the 1950s would be in his seventies, and would be a closer test — not the brother's Y but the man's own. Rows 14 and 31 have both been waiting on one ageing man. STATED CAREFULLY: nobody knows whether there were children of that marriage. The archive has never seen a birth, and eleven years without issue is perfectly ordinary. But the schedule is the cheapest place to find out and until now nobody had a reason to look. Her estate is post-1976 so it is at the Master, not in NAAIRS; order it alongside request 19's three sisters.

NEW, out of row 83, and it belongs in the same Pretoria visit as rows 24, 80 and 91. Marjorie Eileen Dunlop, born Walker, divorced Robert John Dunlop in 1949 and married Leonard in October 1951. A divorce summons of the period recites the date and place of the marriage, the parties' ages and domiciles, AND THE CHILDREN BY NAME AND BIRTH DATE — which is exactly what the estate row wants to know before paying for a copy. It would also place her in Johannesburg in 1949, two years before she married Leonard and one year after 1948. Cheap, because the researcher is already going for five other files. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20.

NEW, out of row 74's sweep, and it belongs in the Pretoria bundle with rows 24, 80, 91 and 93. WLD 2225/1953, MORRIS ISAAC SALKOW VERSUS IVONNE (BORN DE FIGUERIDO) — the divorce filed the same year as the 13 February 1953 marriage. A divorce summons of the period recites the date and place of the marriage, both parties' ages and domiciles, AND THE CHILDREN BY NAME AND BIRTH DATE. For a woman whose stated age drifts younger on every form she signs, and whose parents are the central unsolved question of this archive, A PLEADING SHE HAD TO ANSWER IS WORTH MORE THAN ANOTHER CERTIFICATE. It may also be the last South African document that records where she was living, which is the thing nobody has been able to establish between 1953 and her death at Porto. || FOLDED into REQUEST 20 on 16 September — the consolidated Pretoria court-file bundle. Between 15 and 16 September this archive found EIGHT files in one depot, from five separate lines of enquiry, opened as five separate rows. That is how research happens and a terrible way to commission it: one person, one visit, one list, in priority order so it can be cut from the bottom. The bundle is HELD, not sent — the researcher was engaged on 14 September for six Booyzen files and has not replied, and nothing more goes to her until she answers on rates and capacity. This row stays open only as a pointer to request 20.

NEW and URGENT, out of row 62, AND IT IS THE CHEAPEST DOCUMENT ON THE ENTIRE WORK LIST. GRO marriage certificate, HITCHIN, December quarter 1939, VOL 3A PAGE 4416 — Edward J Bailey and Jeanette Eileen Maud Wear. An English marriage certificate names BOTH FATHERS, with their occupations. IF IT NAMES HENRY ROBERT WEAR, this archive has documentary proof from an official record that he had a daughter born 1919 at Hitchin — and the seven-person MyHeritage tree stops being the sole authority for the couple at the head of this line. It does not prove Leonard was their son; nothing yet does. But it is exactly what requests 19 and 20 are trying to achieve the slow and expensive way, for about £14 and an email. BUY THIS BEFORE ANYTHING ELSE ON THE LIST, including the GRO birth and marriage certificates of rows 27 and 28.

NEW, 16 September, AND IT IS NOW THE SINGLE MOST IMPORTANT ACTION ON THE LIST for the question this archive exists to answer. Four instruments have each looked in a different place and found no death for Leonard Aubrey Robert Wear: the South African estate index (to end 1976), the Johannesburg cemetery registers, the England and Wales death index (1970-1985, >99%, printing dates of birth), and the British consular registers, which structurally cannot hold a South African death of either candidate year. THERE IS EXACTLY ONE PLACE LEFT: THE SOUTH AFRICAN DEATH REGISTER, held by the DEPARTMENT OF HOME AFFAIRS. GenDatabase's 1979 entry is a transcription of that register and is therefore not one source among several but the only source that could exist — and it is flagged INCOMPLETE with the date of birth printed as the corrupt string 0017. WHAT TO ASK FOR: the death register entry (form BI-1663 / the death certificate with full particulars, NOT the abridged version) for WEAR, Leonard Aubrey Robert, died circa 1979. THE FULL CERTIFICATE NAMES THE DECEASED'S PARENTS. So this single document would settle the death year AND test the parentage in one stroke, which nothing else on this list does. BEFORE WRITING: verify the Home Affairs address the way row 48 requires, and check the sent mail — two archives share this mailbox. STRENGTHENED 17 SEPTEMBER. A fifth and sixth instrument have now looked and cannot reach him: FamilySearch's South African civil death index stops in 1966, and the Transvaal estate-file death notices - which would name his parents - are published for 1955 to 1960 only. The claim that the Home Affairs death register is the ONLY PLACE LEFT now rests on six checks rather than four, and each of them is a bound that was measured rather than a search that came back empty.

REWRITTEN 16 September after David pointed out that the first draft answered only one of her questions. SHE ASKED THREE THINGS AND THE REPLY NOW ANSWERS ALL THREE IN HER OWN ORDER, IN THE FIRST SIXTY WORDS. (1) "If you would like me to proceed, I can first check TAB, WLD 2545/1950" — YES, PLEASE DO. That is the file in which IVONE IS THE PLAINTIFF, and knowing the grounds and the outcome is worth more than photographs of the others. (2) "Direct your archival enquiries to me directly and keep the correspondence in one thread" — YES, agreed and apologised for. (3) "If you would prefer that the general enquiries address also be copied, I am happy to do so" — NO, SHE ALONE. She offered that for David's convenience; declining removes a step from her day and is the stronger form of the commitment she asked for. THE REPLY ALSO ACKNOWLEDGES THAT SHE ANSWERED HIS OWN QUESTION — court files are bound like the estates — which he had asked twice, at 03:11 and again at 04:11 on 15 September; without acknowledgement she would reasonably wonder whether he read either message. WHAT IT DELIBERATELY DOES NOT DO IS HAND HER ANOTHER LIST. She offered to check ONE file — "I can FIRST check" — and this archive has eight more TAB files it would dearly like read. Sending them in the same message as an apology for sending too much would be absurd. Accept the one, and let her come back. ON THE DEAD DOMAIN, the letter says only what is provable: enquiries@dac.gov.za bounced with "the recipient's domain does not exist", and there has been NO REPLY from the others on that domain. IT DOES NOT CLAIM THEY BOUNCED — no NDR for archives@ or naairs@ was ever received, which is odd on the same domain and cannot be explained from here. REVISED AGAIN 16 September. David asked what Ms Xaba actually wants — separate enquiries, or one enquiry listing everything? Her own words settle it: she objects to THE SAME ENQUIRY going to several addresses, not to several enquiries; she writes of "your current ENQUIRIES" in the plural; and she wants them in one thread SO THAT SHE CAN PROVIDE A CONSOLIDATED RESPONSE, which is only possible if she holds the whole picture at once. The draft's promise of "one enquiry at a time, and I will not open another" was an over-correction that would have produced MORE separate replies from her, not fewer. THE LETTER NOW CARRIES THE FULL MANIFEST: all fourteen Transvaal references in two tables, one for the Luwinski/Figueiredo marriage and one for the Wear family, each row with a status — nothing needed, bound and understood, with the researcher, or, for WLD 2545/1950 alone, yes please. It is a manifest, not an order form: no priorities, no fee question, no "please supply", and one sentence above it saying plainly that the only thing being asked of her is 2545/1950. It also tells her the private researcher she recommended is doing eight of them on 22 September, which is the courtesy of showing her that her advice was taken. AMENDED AGAIN 16 September, after David asked whether the letter had been updated for the naturalisation finding. IT HAD NOT, AND IT SHOULD HAVE BEEN. The reason for holding request 24 back was that reply 21 promises 'the only thing I am asking of you is WLD 2545/1950' — but THAT REASONING ONLY APPLIES TO THINGS ADDED AFTER SENDING, AND THE LETTER HAS NOT GONE. Two things changed after it was drafted: the naturalisation angle, and NAAIRS going down. Both now sit in the letter. It tells her NAAIRS has lost its DNS record and returns CICS 400 on every path including its front page — offered as information, not a complaint, because it is NARSSA's own system and she may not have been told; it is the only thing in the letter that is a gift rather than an ask. And it asks ONE question that needs no file pulled: is there an index entry under FIGUEIREDO or MOTA FIGUEIREDO in the naturalisation or alien-registration series at SAB. The promise sentence was amended from 'the only thing I am asking of you' to 'THE ONLY FILE I AM ASKING FOR' so that it stays literally true. The question carries its own off-ramp — 'if the answer is simply nothing under that surname, that is a genuinely useful answer and I will stop looking down that road' — and it points her at the father rather than the daughter, with the reason, so she is not sent hunting for a file the law of the day may have made unnecessary. EVERYTHING ELSE IN REQUEST 24 STAYS HELD. READ END TO END 16 September before sending, and TWO THINGS WERE WRONG. (1) THE NATURALISATION PARAGRAPH WAS GARBLED — an earlier edit had spliced two versions and left 'And it is more likely to be her father than her' stranded AFTER the closing sentence, so the paragraph ended mid-argument. Rewritten. (2) IT STILL SAID 'from what I can see on NAAIRS', which the Oppenheimer control had overtaken by then. It now tells her plainly that the online index does not list naturalisation files at all — with the proof, a man whose application is printed in the Gazette of 20 February 1942 and whose file is not indexed — which turns a vague request into a specific reason for asking HER rather than looking himself. ALSO REFORMATTED FOR EMAIL: the markdown pipe tables would have arrived as pipes and dashes in Outlook, and the backticks round dac.gov.za would have printed literally. Plain-text send version at requests/21-SEND-THIS-to-mavis.txt. SENT 16 SEPTEMBER 2026, message 1a0a9d94c42e1756, inside thread 1a0a30c59c9c8405. Sent to her ALONE, not reply-all — she had copied Enquiries2 on her own message, and a letter that declines the cc should not arrive carrying one.

DRAFTED as REPLY 22, not sent, AND IT IS TIME-CRITICAL. She replied on 15 September with a quote and a date: R2,850 for the first file INCLUDING THE DRIVE TO PRETORIA, then subsequent files charged by the work in them, "not to exceed R2,000" each — the sentence is garbled and the ceiling should be confirmed rather than assumed. No per-page or hourly charge. Invoice before the visit. She confirms MHG 52653 is a normal reference despite the missing year. AND SHE OFFERS TUESDAY 22 SEPTEMBER. HER COST STRUCTURE IS THE WHOLE ARGUMENT: the first file carries the journey and every later file is cheap BECAUSE SHE IS THERE ANYWAY. Request 20 holds eight more TAB files and the Lerena archive has at least five. EVERY FILE ADDED TO THIS TRIP IS CHEAP; EVERY FILE THAT MISSES IT COSTS ANOTHER R2,850. The list is not ready, so the reply confirms the six, accepts the date, and asks one question: by when does she need the additional references to include them on the 22nd. REWRITTEN ENTIRELY 16 September after David asked whether the Anne and Mavis letters were up to date. MAVIS'S WAS. ANNE'S WAS NOT — its central premise had been overtaken by two emails nobody had read. HER TERMS, 15 Sep 20:02: NO per page, NO hourly, NO call-out; R2,850 FOR THE FIRST FILE including the drive to Pretoria and back; each file after that charged by how much work it took, NOT TO EXCEED R2,000; 'I would not exploit the situation in any way.' She capped her own downside unprompted. INVOICE FIRST, THEN SHE TRAVELS. HER MESSAGE OF 16 Sep 08:46: the invoice is coming, it covers THE SIX MHG FILES PLUS A SEVENTH FREE — MHG 10245/69, the Wear death notice, 'as I can see it is just one page, so I am giving it to you for nothing' — and SHE GOES ON TUESDAY 22 SEPTEMBER IF THE INVOICE IS ACCEPTED. THE OLD DRAFT'S 'SCALE PROBLEM' IS DEAD. It agonised over 26 items against 6 quoted; she has settled the scope herself — 'So lets do the 6, and it will go well' — and has ALREADY AGREED IN PRINCIPLE to the CJC Marico files, the OD series and the Bloemfontein estates: 'I can do this for you.' Raising a scale problem now would solve a problem she has already solved. The tiered appendix is HELD for the second visit. AND SHE ASKED ONE DIRECT QUESTION WHICH IS THE ONLY THING HOLDING HER UP: 'For the ANC at the deeds office, we need the name and the date. (you did give me the dates,)' SHE IS RIGHT AND IT IS THIS ARCHIVE'S OMISSION — the 03:25 email of 15 September gave her two dates and NAMED NOBODY, and a deeds registry indexes ante-nuptial contracts by the parties. The reply supplies all three with names: 19 JUNE 1948, KURT ISRAEL LUWINSKI and IVONE born MOTA FIGUEIREDO formerly DE WITT; 29 FEBRUARY 1956, register entry No. 169, KURT ISRAEL LUWINSKI and FRANCES HOSTOWSKY; and 30 OCTOBER 1951, certificate 1397, LEONARD AUBREY ROBERT WEAR and MARJORIE EILEEN DUNLOP born WALKER — THE THIRD IS NEW TO HER AND IS THE ONE TO GET, because it is the only one that names Leonard and an ANC recites each party's full particulars rather than a register's eleven columns. DAVID'S ACTIONS IN ORDER: wait for the invoice and pay it, because she does not travel until it is settled and Tuesday is six days off; send the reply so the names reach her first; do NOT send the appendix. REFORMATTED FOR EMAIL 16 September. The three group tables were markdown pipe tables and would have arrived in Outlook as a wall of pipes and dashes — for a woman about to work from them at a reading-room desk, that is not a cosmetic problem. Rewritten as indented numbered items with the reference on one line and the want beneath it, which also survives printing. Plain-text send version at requests/22-SEND-THIS-to-anne.txt. SENT 16 SEPTEMBER 2026, message 1a0a9d8a6c8c5783, inside her existing thread 1a0a234a65602bca. WHAT IS NOW OUTSTANDING IS DAVID'S, NOT THE ARCHIVE'S: she does not travel until the invoice is settled and Tuesday is six days off. AND A CONSEQUENCE TO EXPECT — the letter widens the job from seven items to twenty-six and explicitly invites a revised invoice, so a larger number may come back before Tuesday. That is the price of one drive instead of two and it was taken deliberately.

FOUND AND FIXED 16 September, and it is the most serious error this archive has made. The standing rule is that LIVING DNA MATCH HOLDERS ARE NEVER NAMED — role, country and cM only. They never consented to anything and their names reached this archive solely because they share chromosomes with Derrick. FOUR WERE PUBLISHED IN FULL on /worklist/ and FIVE MORE BY SURNAME on /dna/, in a sentence identifying them as people rather than as ancestral surnames. It was found only because the AutoClusters export arrived and gave, for the first time, an authoritative list of all 111 match holders to check the build against. NOTHING CAUGHT IT BECAUSE NOTHING WAS LOOKING: check-living guards the people in the family REGISTER and says "none present" while four match holders sat on the work-list page. Both pages are corrected and rebuilt. THE DURABLE FIX IS tools/check-matches.py, now wired as npm run check:matches: it cross-references every name in the export against every built page, flags full names always and surnames unless they are on a documented AMBIGUOUS list — Halsey and Morris and Brown are there because they are also DEAD people in this archive, with the reason written beside each. It degrades gracefully: with no export on the machine it SKIPS and says so, rather than giving a false all-clear. Verified by injecting a name into the build and confirming it fails.

NEW, out of the AutoClusters report, AND IT OUTRANKS ROW 44's CANDIDATE. Cluster 3 — the paternal cluster — holds twelve people and 1,706 cM, nearly a third of all clustered sharing in the kit. Eight of the twelve HAVE NEVER APPEARED IN THIS ARCHIVE, and one of them shares 165.9 cM, WHICH MAKES HIM THE FOURTH-LARGEST MATCH DERRICK HAS. At that level he is closer than a second cousin. Run his segments against data/wear-signature.json: if most fall in the twelve regions the Castle cousins do not share, the Wear line has a second living witness and row 31's Y-DNA problem gains another candidate to approach. Needs the chromosome browser, so it needs the sign-in. DO NOT NAME HIM ANYWHERE IN THE SITE.

DRAFTED as REQUEST 23, HELD — these go to living private individuals and only David can send them, which row 6 already establishes. THE CASE FOR THEM IS THE STRONGEST THING IN THE AUTOCLUSTERS REPORT. Four of the twenty clusters are Portuguese — fifteen people, about 776 cM — and because Derrick's mother was Portuguese from Lourenço Marques, EVERY ONE OF THEM IS ON HER SIDE. Three separate clusters means three separate ancestral couples, which is what a Figueiredo–Mota marriage plus their own parents' lines should look like; one is probably the MOTA line and another the FIGUEIREDO. THE SCALE, PLAINLY: the strongest shares 133.2 cM, MORE THAN ANYONE THIS ARCHIVE HAS EVER WRITTEN TO ABOUT IVONE, and more than the letters to Lisbon, Maputo and Porto have produced between them. The letters are in Portuguese with English below, sent through MyHeritage's own messaging because the export gives tree links and not addresses. ONE MESSAGE PER CLUSTER, to the strongest member, saying openly that one or two others in the same group are being written to — NOT all fifteen, who may well know each other and would read fifteen near-identical messages as harvesting. Order: the 133.2 cM cluster, then the one with two over 100, then the one whose members share an 8,911-person tree, then the weakest only if the others produce nothing. WHAT IS DELIBERATELY ABSENT: anything about the paternity question. It is Derrick's private business, it concerns his FATHER's side, and it has no bearing on what a Portuguese cousin can help with. The maternal line is not in doubt, and a stranger told a delicate thing in a first message usually stops replying. The ask is one question — do the names MOTA, FIGUEIREDO or PINTO DA ROCHA appear in your family, and is there any link to Mozambique — with an explicit statement that no money or estate is involved. THE FIRST USEFUL ANSWER IS A PLACE: a parish or district in Portugal. With a parish the Portuguese civil registers become searchable and row 65's dead end reopens at no cost.

RAISED 16 September, when David asked whether anything from the Booyzen and Lerena sessions ought to go to Ms Xaba rather than to Anne. CHECKING THE ASSIGNMENT SHOWED ALL 26 ITEMS POINTED AT ANNE, AGAINST A SIX-ITEM BOOKING — and several of them are the wrong shape for a paid research day. THE THING THAT CHANGED IS BIGGER THAN THE LIST: since 14 September this archive has read "we are unable to SCAN these files" as "bound files are unreachable". She never said she could not READ them, and her offer on WLD 2545/1950 proves she will look at a file and say what it is in. A J294 death notice is wanted for one thing — who does it name as parents, spouse, children — and that is a sentence. This archive has been commissioning photographs of sentences. THE QUEUE IS RANKED BY HOW BADLY AN ITEM SUITS A RESEARCHER. Q1: the two doubtful references, TAB MHG O/2899 and MHG 52653, which Anne is being told to skip — a bad reference is a desk question, and Ms Xaba has already demonstrated exactly that skill by finding WLD 2545/1950 where NAAIRS spells the surname Luwenski. Q2: whether the OD correspondence-register numbers pull as discrete items or sit inside bound volumes — one sentence from her decides whether Anne spends ten minutes or ninety. Q3: CJC vol 907 ref 104 and SAB VWR 2 B645/8/38, where the index line is the entire catalogue record. Q4: four death notices, held back deliberately because a photograph beats a summary and Anne is already going — they are the fallback if her day runs out, so that the answer is a question rather than a second invoice. NOTHING GOES UNTIL SHE ANSWERS ON 2545/1950. Reply 21 told her the only thing being asked of her was that file, and breaking it in the next message would cost more than the whole queue is worth. When it does go, Q1 and Q2 travel together in one message — they are all 'does this exist and how does it pull' — which is the consolidation reply 21 promised.

RAISED 16 September, from David's question: a woman naturalised out of Lourenço Marques would have a file naming her parents, wouldn't she? THE RECORD CLASS IS REAL AND THIS ARCHIVE HAS NEVER TESTED IT. A South African naturalisation memorial asks for date and place of birth, THE FATHER'S NAME AND NATIONALITY, the date and ship of arrival, and the wife and children by name and age. AND THE FATHER'S FILE WOULD BE BETTER THAN HERS — it names every child, so it would give her parents AND her siblings in one document, which nothing this archive has tried can do. BUT THE LAW MAY HAVE REMOVED HER OWN FILE BEFORE IT EXISTED. An alien woman who married a British subject before 2 September 1949 acquired his nationality automatically; the South African Citizenship Act 44 of 1949 ended that and required an alien wife to naturalise like a man. IVONE MARRIED JOHANNES ANTONIE OELOFSE DE WITT ON 2 FEBRUARY 1948, nineteen months inside the old rule. If de Witt was a Union national she became one by marriage and there is no file for her — AND NATIONALITY ACQUIRED BY MARRIAGE SURVIVED THE DIVORCE, so TPD 534/1948 does not undo it. THAT MAKES DE WITT'S OWN NATIONALITY A QUESTION WORTH ANSWERING, because it decides whether her file can exist. THE GAZETTE HAS ALREADY BEEN SWEPT: 1930 to 1960, every year, surname Figueiredo, South African issues only — no naturalisation notice, with the control test run first and passing. On /negatives/. NAAIRS, which indexes the SAB naturalisation files, WAS UNREACHABLE on the day. The enquiry is queued for Ms Xaba as Q5 of request 24, framed as 'is there an index entry for the surname' rather than 'find her file'.

FOUND 16 September as a by-product of the naturalisation sweep, and neither was known to this archive. ESTATE 7029/54 — FIGUEIREDO, AGOSTINHO VICENTE SEBASTIAO, of Johannesburg, first and final account advertised 22 July 1955. ESTATE 2910/60 — MONTEIRO, JOSÉ DE FIGUEIREDO, OF LOURENÇO MARQUES, PORTUGUESE EAST AFRICA, executor Frederick William Damant c/o Standard Bank, advertised 2 December 1960. THE SECOND IS THE INTERESTING ONE: a man of Lourenço Marques whose estate was administered in the Union, in the decade after Ivone's marriages, which is exactly the profile of somebody the family might have known or been. NEITHER IS CLAIMED AS A RELATIVE — Figueiredo is a common Portuguese surname and 'Monteiro, José de Figueiredo' is a different name from hers. What makes them worth a row is that each has an estate number, and a South African estate file holds a death notice, form J294, which asks for the deceased's parents. THE CHEAP NEXT STEP IS THE GAZETTE, NOT THE FILE: an estate is advertised more than once, and the earlier notices carry the date of death and often the address.

RAISED 16 September on David's suggestion, and it is the kind of thing that decides whether a search works. IVONE IS RECORDED UNDER SIX SPELLINGS OF HER MAIDEN SURNAME ACROSS FIVE DOCUMENTS, and NAAIRS itself carries two of the worst. FORENAME: IVONE in most files, YVONNE in TPD 534/1948, IVONNE in WLD 2225/1953. MAIDEN SURNAME: FIGUEIREDO and DE FIGUEIREDO; MOTA FIGUEIREDO in WLD 1950/1948; DE FIGUERIDO in WLD 2225/1953; DE FIGUEREDO on her own 1953 marriage register; and FIGARETTO in the NAAIRS index. MARRIED SURNAMES: DE WITT from February 1948, LUWINSKI — also LUWINSKY and LUWENSKI — from June 1948, SALKOW from February 1953. THE PROOF THAT IT MATTERS IS THE FILE MS XABA HAS OFFERED TO CHECK: WLD 2545/1950 hid from this archive for weeks because the index spells her LUWENSKI. The table now goes into reply 21, framed as saving her a search rather than instructing her. IT ALSO COST THIS ARCHIVE A NEGATIVE: the gazette naturalisation sweep had searched only three of the six spellings. RE-RUN THE SAME DAY across all six plus the married names — no naturalisation notice under any of them, and FIGARETTO, FIGUEIRIDO and LUWENSKI turn out to appear NOWHERE IN ANY AFRICAN GAZETTE, which makes them index corruptions rather than spellings anyone ever printed. The negative on /negatives/ is corrected and is now stronger than when it was first written, not weaker.

DONE 16 September on David's instruction: 'can you also make sure the site is clear on the different spellings'. NEW PAGE /names/, in the primary nav after Ivone, with data in site/src/data/names.json. FOURTEEN FORMS, each given with the record it comes from and marked as either a document variant or an index corruption — a distinction that matters, because an index corruption tells you about the cataloguer and a document variant tells you about the family. Three forenames: IVONE, YVONNE (TPD 534/1948), IVONNE (WLD 2225/1953). Six maiden surnames: FIGUEIREDO, DE FIGUEIREDO, MOTA FIGUEIREDO (WLD 1950/1948, the fullest anybody wrote and the only record giving her a second family name), DE FIGUERIDO (WLD 2225/1953), DE FIGUEREDO (her own 1953 marriage register, written by a clerk with her standing there), FIGARETTO (the NAAIRS index). Five married: DE WITT, LUWINSKI, LUWINSKY, LUWENSKI, SALKOW. THE PAGE LEADS WITH WHAT IT COST: WLD 2545/1950 is the only action in which she is the plaintiff and it went unfound for months because the index carries her as IVONE LUWENSKI BORN FIGARETTO. It also states the rule now worked to — no surname search counts as done until every recorded spelling has been tried — AND THE COUNTER-RULE, that FIGARETTO and LUWENSKI appear nowhere in any African gazette, so a spelling that finds a file can still be one nobody ever wrote: useful for searching, useless as evidence. Cross-linked from /ivone/, and the other lines' variants (Kurt; Wear/Weir/Ware; Booyzen/Booysen/Booijzen/Booijsen) listed so the problem is not treated as hers alone.

RAISED AND PART-DONE 16 September: 'all those spelling differences should now go and research everything we have to-date with the different variations'. DONE SO FAR. (1) UNION GAZETTE, 1930-1960, every year, six maiden spellings plus the married names — no naturalisation notice under any of them; FIGARETTO, FIGUEIRIDO and LUWENSKI return NOTHING ANYWHERE IN ANY AFRICAN GAZETTE. On /negatives/. (2) GENDATABASE, twelve name forms fully paged — 27 National-Archives rows that the archive's own naairs-figueiredo.tsv does not contain. The haul includes SEVEN IMMIGRATION-PAPERS FILES under Figueiredo and Figueredo, all at KAB in the IRC and PIO series, none of them hers but PROVING THE RECORD CLASS EXISTS AND IS INDEXED; and MORRIS ISAAC SALKOW'S ENTIRE LITIGATION HISTORY — TPD 467/1950, TPD 2854/1960, WLD 1593/1962, WLD 3317/1963, WLD 15934/1979, and his estate papers at KAB MOOC 6/9/5171 61416 — for the man Ivone married in February 1953 and had left by the end of it. Also OPPOSED APPLICATION. PAULA LUWINSKI VERSUS PHILLIP CRYSTAL, TAB WLD 2451/1972, a Luwinski forename this archive has not seen. (3) NAAIRS, on the corrected endpoint: FIGUEIREDO 43, FIGARETTO 1, LUWENSKI 1, LUWINSKY 1 — each variant surfacing a file the correct spelling does not return. STILL TO DO: FamilySearch, now signed in again, on all fourteen forms; the remaining ten spellings against NAAIRS; and the Salkow and Paula Luwinski finds worked up properly rather than logged. COMPLETED 16 September for NAAIRS and gendatabase. FORTY-ONE SPELLINGS TESTED — twenty-one ways of writing Figueiredo and twenty of writing Luwinski, INCLUDING MANY INVENTED ON PURPOSE to see whether a clerk had ever made the same slip. ONE GENUINELY NEW RECORD, AND IT IS A GOOD ONE: TAB MHG 7029/54, the Johannesburg estate of a Portuguese man the Union Gazette advertises as FIGUEIREDO, AGOSTINHO VICENTE SEBASTIAO and NAAIRS files as FIQUEIREDO — WITH A Q. Two indexes, two spellings, one man, and nobody had ever typed the Q. Also KAB PIO 990/42707E, IMMIGRATION PAPERS JG DE FIQUEIREDO, and SAB BAL 7, a temporary permit for a Goanese C FIGUERDO — which together prove the permit and immigration series are indexed and reachable. HE IS NOT CLAIMED AS KIN: Figueiredo is common and nothing connects him to Ivone but a city and a decade. What his file carries is a death notice, and a death notice names parents. AND ONE REAL EXCLUSION, worth as much as a find. Kurt came out of East Prussia and the slide from LUWINSKI to LEWINSKI is exactly the slide a clerk makes; there is a Lewinski and a Lewinsky family in Johannesburg in the same decades — Erwin, Bella born Berman, Meinhard, Ingeborg Ursula born Leske. THEY ARE NOT THIS FAMILY, and that is now recorded so it is never re-opened as a hopeful possibility. Seventeen invented spellings returned nothing anywhere, which is itself the finding that THE LUWINSKI NAME CORRUPTS IN ONLY TWO DIRECTIONS, NOT TEN. All of it is on /names/.

DAVID TOOK A ONE-MONTH FULL SUBSCRIPTION ON 16 SEPTEMBER 2026 AND OPENED IT IN THE BROWSER. THE CLOCK IS THE POINT: everything that needs Findmypast should be done inside the window, and this row exists so that the window is not discovered to have closed. IT IS THE BEST INSTRUMENT THIS ARCHIVE HAS EVER HAD FOR THE WEAR LINE, because that line is entirely English and Findmypast holds what FreeBMD and the GRO index do not. PRIORITY ORDER. (1) PASSENGER LISTS LEAVING THE UK, BT27, 1890-1960 — THE SINGLE HIGHEST-VALUE TARGET IN THE WHOLE SUBSCRIPTION. The Wear family emigrated entire and this archive has never dated it. An outbound list gives the ship, the date, the destination port, the ages and THE WHOLE PARTY TRAVELLING TOGETHER, which would settle in one document whether Leonard went with his parents or followed. (2) THE 1939 REGISTER, which carries EXACT DATES OF BIRTH, not just quarters. If Leonard is absent from it he had already gone by September 1939; if he is present, it dates his departure to between then and his Rosebank marriage in September 1940. Either answer is worth having, and the register would also give Henry Robert and Martha Maud's exact dates. (3) BRITISH NEWSPAPER ARCHIVE — Hitchin and Letchworth titles, for Wear and Castle notices: births, marriages, deaths, court, emigration send-offs. (4) ELECTORAL REGISTERS for 72 Pix Road, Letchworth, and the years either side, to bracket the departure from the other end. (5) PARISH REGISTERS AND BAPTISMS for the six Hitchin children 1917-1936, which may name godparents and addresses the GRO index cannot. (6) THE CASTLE LINE — Martha Maud Castle's own origins, still thin. (7) PROBATE AND WILLS. HARVEST RATHER THAN BROWSE: save what is found into data/ as it goes, because the subscription ends and the evidence should not. STOCK-TAKEN 20 SEPTEMBER, FOUR WEEKS BEFORE THE WINDOW SHUTS, AND THE SEVEN PRIORITIES CHECKED RATHER THAN ASSUMED. (1) BT27 IS DONE and was worth the subscription on its own — three sailings, three dates, in data/wear-emigration-bt27.tsv. (2) THE 1939 REGISTER IS DONE. (5) is done for the Castles and NOT for the six Hitchin Wear children, which is a real remaining gap. (6) is largely done. (4) ELECTORAL REGISTERS are now largely redundant: three passenger lists fix the departure to the day from the other end. (7) PROBATE belongs with the English estate work. (3) THE NEWSPAPERS WERE TRIED AND THE INSTRUMENT WILL NOT DO IT. "WEAR" IS ONE OF THE COMMONEST WORDS IN BRITISH NEWSPAPER ADVERTISING and the newspaper name box is an OCR string match, not a name index like the census sets. Wear + Letchworth returns 9,442 articles headed "MEN'S WEAR AT PUGH'S, Letchworth"; "Martha Maud Wear" returns 285 headed by a knitting joke and a sock advertisement. TWO FURTHER PROPERTIES OF THE TOOL, WORTH THE SAME WEIGHT AS THE YEAR-FILTER RULE: the EXACT-KEYWORD switch DOES NOT ENFORCE A PHRASE — "Pix Road" with it ticked returns 85,719 — and the RESULTS URL IGNORES the county and date parameters, so this set cannot be swept by constructing URLs the way the record sets can; a query with all five returned 20,426,315, more than the unfiltered one. RE-SCOPED RATHER THAN ABANDONED: a newspaper sweep for this family must be driven by strings that are not ordinary English — SHENTON, GIDDINS/GIDDINGS, WALLIS with Baldock, HULL with St Albans — and by the ship names WINDSOR CASTLE (March 1937) and DUNVEGAN CASTLE (August 1939) against the Hertfordshire titles, for a departure notice. In data/findmypast-newspapers.tsv. THE SURNAME WEAR ON ITS OWN IS NOT WORTH TRYING AGAIN. PRIORITY 5 WORKED FOR THE WEAR CHILDREN THE SAME DAY, AND IT IS A NEGATIVE PLUS A DEFECT. HERTFORDSHIRE BAPTISMS HOLDS ELEVEN WEAR ROWS IN THE WHOLE COUNTY — 1570 Watford, 1593 Kings Walden, 1594 Little Hadham, 1670 Berkhamstead, 1786 HITCHIN, 1845 St Albans St Michael twice, 1884 Little Wymondley, 1891 Leavesden three times. THE LATEST IS 1891 AND THE ONLY HITCHIN ENTRY IS 1786. None of the six children is in it, and a whole-category search for CYNTHIA Y E WEAR, born Hitchin 1931, returns seven results of which every one is a CIVIL index — no parish baptism anywhere on Findmypast. Whether that is a coverage bound or a real absence is NOT established: the Castle work measured Baldock ending in 1879, which points to a bound, and the set's own coverage page 404s. The route is Hertfordshire Archives or the parish, not this subscription. AND THE DEFECT MATTERS MORE THAN THE ANSWER. The rule from 16 September — that Findmypast's MARRIAGE index silently returns zero when year-filtered — IS NOT ONLY THE MARRIAGES. Hertfordshire Baptisms, lastname=smith, no year: 15,122. The same search at year=1880 give-or-take 5: ZERO. At 1920: zero. Without the 1880 control the 1920 zero would have been written down as a coverage bound. THE RULE RESTATED WIDER AND RECORDED IN data/castle-baptisms-weston.tsv: NEVER YEAR-FILTER A FINDMYPAST SEARCH IN ANY SET, and if one is run, run a positive control in the same set in the same breath. PRIORITY 7, PROBATE, WORKED THE SAME DAY AND IT PAID. GEORGE WILLIAM CASTLE THE YOUNGER — Martha Maud's elder brother, the second of the nine — has been carried in this archive on a date taken from A MATCH HOLDER'S USER TREE and nothing else: "abt 1885 Baldock, d. 23 Aug 1939". The England & Wales GOVERNMENT PROBATE DEATH INDEX has him, record GBOR/GOVPROBATE/C/1939-1939/00022564: DEATH DATE 23 AUG 1939, RESIDENCE TOWN LETCHWORTH, probate granted 2 NOVEMBER 1939 at the LONDON registry, LINKED PERSON CYRIL GEORGE CASTLE. The user tree's date is confirmed to the day by a government index. TWO THINGS COME WITH IT. (a) HE WAS IN LETCHWORTH — the Wears' town, and the youngest Castle brother was at 9 Boscombe Place, Letchworth in 1921. The Castles had moved from Baldock to Letchworth and the Wears went to the same place; stated as a fact about addresses and not offered as a reason for anything. (b) THE DATES SIT THIRTEEN DAYS APART: Martha Maud sailed from London on the Dunvegan Castle on 10 AUGUST 1939 and her brother died at Letchworth on 23 AUGUST 1939. Both halves documented, BT27 on one side and the probate index on the other. AND A NEW NAME: CYRIL GEORGE CASTLE, the linked person, who is NOT the ninth child Ernest Albert Cyril Castle of Baldock. Likeliest a son of the deceased; not asserted. NOT READ: the original calendar page, which would give the street address, the occupation, the executors and the estate value. It is IIIF at 2864x4361 behind Findmypast's image gateway and the gateway is SESSION-LOCKED — a request carrying the visible cookies returns 403. Record id GBOR/GOVPROBATE/1939/C000740-CASSON-1939 for whoever opens it next in a signed-in browser. In data/castle-nine-children.tsv. PRIORITY 3 RE-RUN AND CLOSED, 20 SEPTEMBER, ON A COVERAGE BOUND. The re-scoping said to drive the FILTER PANEL rather than the address bar and to search strings that are not ordinary English. Both were done. FIRST, WHY THE PANEL WORKS NOW: THE BROWSER PANE HAD BEEN HIDDEN, and a click cannot reach a pane that is not compositing frames — so every earlier attempt to open a facet failed with the panel apparently inert. Displaying the pane fixed it instantly. The same hidden pane caused the Szubin padlock error an hour later. Twice in one day, one cause. THE SHIP WAS THE BEST IDEA IN THE RE-SCOPING AND IT CAME BACK EMPTY: keyword "Dunvegan Castle", 47,783 articles, faceted to HERTFORDSHIRE gives ELEVEN — 1890, 1897 x2, 1898, 1900, 1901 x2, 1902, then 1982 x2 and 1986. NOT ONE IN 1939. AND THEN THE COVERAGE WAS CHECKED, WHICH SHOULD HAVE COME FIRST. LETCHWORTH HAS EXACTLY ONE DIGITISED TITLE — CITIZEN (LETCHWORTH), 1906-1917, 587 ISSUES. The Wears left in 1937-39. THE TOWN'S ONLY PAPER STOPS TWENTY YEARS BEFORE THEY WENT. HITCHIN HAS EXACTLY ONE — HERTFORDSHIRE EXPRESS, 1859~1956, 1,403 ISSUES, and the tilde is Findmypast's mark for an incomplete run: fourteen issues a year for a weekly is a quarter-sampled run at best. SO PRIORITY 3 IS CLOSED AND NOT MERELY RE-SCOPED. A departure notice for this family is not findable here, and the reason is neither the search nor the surname — the town's paper was never digitised for the years the family was leaving. What would still work is reading-room work: the British Library holds the Citizen beyond 1917 on microfilm, and Hertfordshire Archives at Hertford holds local runs. Neither ordered nor asked for. In data/findmypast-newspapers.tsv. 12 OCTOBER 2026 — THE LAST UNFINISHED ITEM IN THIS ROW IS DONE, FOUR DAYS BEFORE THE WINDOW SHUTS. Priority 7 left one thing unread: the probate CALENDAR page for George William Castle the younger, session-locked behind the image gateway. David signed the session in and it was read in four overlapping native-resolution regions. It gave the STREET ADDRESS (85 Jackmans-place), the PLACE OF DEATH (St Bartholomew's Hospital, London E.C.1 — SO HE DID NOT DIE AT LETCHWORTH, which this archive had said), ADMINISTRATION RATHER THAN PROBATE so there was no will, A WIDOW NOBODY HAD (Margaret Castle), Cyril George Castle's OCCUPATION (printer), and the estate at £1166 13s. 2d. THE ROUTE IS RECORDED FOR NEXT TIME: the record page needs BOTH ids — search.findmypast.com.au/record?id=<IMAGE id>&parentid=<TRANSCRIPT id>; the image id alone returns 500 and /transcript? with the image id returns 404. AND TWO MORE PROPERTIES OF THE TOOL, IN THE SAME FAMILY AS THIS ROW'S YEAR-FILTER RULE: page= IS NOT THE PAGINATION PARAMETER AND DOES NOT FAIL LOUDLY — &page=5 returns “0 results found” because an unknown parameter is treated as a filter nothing matches; it is _page=. And THE PLACE PARAMETER DIFFERS BY SET: keywordsplace= filters the 1841 census and does nothing at all to Hertfordshire Baptisms, where place= is the one that works. WHAT IS STILL LIVE IN THE WINDOW: the Aldenham parish sets — “William Hawkins × Mary, 1800” for Mary's maiden name, the LARKIN generation, and the 1851 census for John and Emma's household. See row 128.

FOUND 16 September by testing FIQUEIREDO, a spelling that appears in no document this archive holds and was tried only because David said to search every variation. NAAIRS: FIQUEIREDO, AGOSTINHO VICENTE SEBASTIAO, ZA TAB MHG 7029/54. The Union Gazette of 22 July 1955 advertises the first and final account in the same estate under FIGUEIREDO. THE TWO INDEXES DISAGREE ABOUT THE MAN'S NAME, which is why he has been invisible. WHAT HE IS NOT: a proven relative. Figueiredo is among the commonest Portuguese surnames and nothing ties him to Ivone except Johannesburg and the 1950s. WHAT MAKES HIM WORTH A ROW: an MHG estate file contains a death notice, form J294, and that form asks for the deceased's parents, his place of birth, and his surviving spouse and children. IF HE IS KIN, IT NAMES THE GENERATION THIS ARCHIVE CANNOT REACH; if he is not, one J294 closes him off for good. HE DIED IN 1954, six years after Ivone's Johannesburg marriages, in the same Portuguese community of the same city. CHEAP FIRST STEP BEFORE ORDERING ANYTHING: the gazette carries estates more than once, and the earlier notice gives the date of death and usually an address. Do that before spending a researcher's hour on it.

FOUND 16 September as a by-product of the spelling sweep, and it changes where to ask. ACROSS EVERY FIGUEIREDO, FIQUEIREDO, FIGUEREDO, FIGUERIDO AND FIGUEIRA SEARCH, THE IMMIGRATION RECORDS CLUSTER IN ONE PLACE: KAB, the Cape Town Archives Repository, in the PIO series (Principal Immigration Officer) and IRC (immigration control). Sixteen of the forty Figueira records sit there; so do JG de Fiqueiredo at PIO 990/42707E, HA Figueired at PIO 33143E, Juonet Figueredo at IRC 1/4/19, Maria T Figueredo at IRC 1/3/1, and five further Figueiredos. NOT ONE OF THEM IS IVONE. WHAT IT CORRECTS: this archive had assumed naturalisation and alien files were central-government records at SAB in Pretoria, and the draft to Ms Xaba asked her for exactly that. A request pointed at the wrong depot comes back empty FOR THE WRONG REASON, which is the most expensive kind of negative there is. THE LETTER IS AMENDED to say what was actually observed and to ask the better question: whether there is a TRANSVAAL equivalent for someone who came in overland from Lourenço Marques rather than by sea to Cape Town. Ivone's route matters here — Mozambique borders the Transvaal, and a woman arriving that way may never have passed a Cape Town immigration officer at all. NEXT: put the variant spellings to the KAB PIO and IRC series directly, and ask the Western Cape Archives, where this archive already has an open request.

FOUND 16 September 2026, within the first hour of the Findmypast subscription, in BT 27 — the outbound passenger lists. THIS ARCHIVE HAS ASKED SINCE THE BEGINNING WHEN THE WEARS LEFT ENGLAND AND HAS NEVER BEEN ABLE TO ANSWER IT. HENRY ROBERT WENT FIRST, ALONE. Mr H R Wear, born 1890, aged 49, occupation MACHINE MINDER (the transcriber read it as 'machine minister'), sailed from LONDON ON 5 APRIL 1939 aboard the Union-Castle liner DURBAN CASTLE for the Cape. TNA/BT27/1540/00/0042 entry 00189. He is the only Wear on that manifest. MARTHA MAUD FOLLOWED FOUR MONTHS LATER WITH FOUR CHILDREN. On 10 AUGUST 1939 the DUNVEGAN CASTLE left London for Durban, 406 passengers, master A Alderson. FIVE WEARS IN FIVE CONSECUTIVE ENTRIES on TNA/BT27/1542/00/0024: 00223 MRS M WEAR b.1895; 00227 MISS K WEAR b.1921 age 18; 00224 MISS C WEAR b.1931 age 8; 00225 MISS L WEAR b.1933 age 6; 00226 MAST B WEAR b.1937 age 2. THE DATE IS THE PART TO SIT WITH: they sailed on 10 August and Britain declared war on 3 September. A month later she would not have gone. AND IT CORROBORATES THE SIBSHIP THIS ARCHIVE BUILT OUT OF THE BIRTH INDEX. The reconstruction said six children at Hitchin 1917-1936 of whom TWO WERE SONS, one thirty-one in June 1948 and the other eleven. MAST B WEAR B.1937 IS THAT SECOND SON, on a government record rather than inferred from a mother's maiden name. MISS C WEAR B.1931 IS CYNTHIA, whose GRO birth entry is what the entire Castle chain hangs from. NEITHER SHIP CARRIES LEONARD. He was 22 in 1939, in Johannesburg by 1940, married at Rosebank 7 September 1940 — so he travelled separately as an adult would. A BT 27 search on the forename Leonard with surname Wear returns NOTHING AT ALL, so he is either indexed under initials or left from a port these lists do not cover. LIVE QUESTION, NOT A CLOSED ONE. Saved to data/wear-emigration-bt27.tsv and published on /leonard/.

TWO THREADS OPENED BY THE BT 27 FIND AND NOT YET PULLED. (1) LEONARD'S CROSSING. He is on neither 1939 ship. BT 27 searched on forename LEONARD plus surname WEAR returns zero. NEXT ATTEMPTS: initials L, L A, L A R; surname wildcards; destination country South Africa with birth year 1917 and a wide tolerance; and the inbound lists if Findmypast holds them, because a man who went out in 1939 or 1940 may appear coming back. THE WINDOW IS TIGHT AND KNOWN: he is in England for the 1921 census at 72 Pix Road and in Johannesburg for his Rosebank marriage on 7 SEPTEMBER 1940. (2) HENRY ROBERT'S MILITARY RECORD. Findmypast surfaced FOUR military records against his transcript — H R Wear 1916 Great Britain, Henry R Wear 1914 Great Britain, H R Wear 1914 Great Britain, and HENRY ROBERT WEAR 1914 SURREY, ENGLAND. He married Martha Maud Castle at Hitchin in Q4 1914. A service record gives a full date of birth, a next of kin and an address, AND IT MAY GIVE HIS OWN FATHER — which is a generation this archive has never reached on the Wear side. FOUND 16 SEPTEMBER, AND IT REVERSES THE STORY. BT 27 on the forename LEONARD returns nothing; ON THE INITIAL L IT RETURNS HIM AT ONCE. TNA/BT27/1490/00/0047 entry 00171: MR L A R WEAR, AGED 20, BORN 1917, OCCUPATION ENGINEER, sailed from SOUTHAMPTON on the Union-Castle liner WINDSOR CASTLE on 12 MARCH 1937, bound for the Cape. THE ORDER OF THE EMIGRATION IS NOT WHAT THIS ARCHIVE PUBLISHED THIS MORNING. It said the father went ahead in April 1939 to find work and a house, and the family followed in August — the oldest pattern in emigration. IT IS THE WRONG WAY ROUND. THE SON WENT FIRST, ALONE, AT TWENTY, TWO YEARS BEFORE ANY OF THEM: Leonard March 1937, Henry Robert April 1939, Martha Maud and the four youngest August 1939, and Janette not at all. AND THE OCCUPATION IS THE SAME WORD THE 1951 MARRIAGE REGISTER GIVES HIM — ENGINEER — which he was already calling himself at twenty, before he left. IT ALSO PUTS A DATE INSIDE A HOLE. Row 41 records that Leonard cannot be placed between 1921 and 1940: 72 Pix Road in the 1921 census, then nothing until Roberts Heights in September 1940. 12 MARCH 1937 NOW SITS IN THAT GAP, and it is the first fact in it. FOUR MILITARY RECORDS were surfaced against the same transcript and have NOT been read — L A R Wear, GERMANY 1939, and Great Britain 1943, 1944 and 1945 — which is a problem as much as a lead, because a man who sailed to the Cape in 1937 and married at Rosebank in 1940 should not obviously be in Germany in 1939. Next row.

FOUND 16 September 2026 on Findmypast, four records in sequence, and it breaks the ceiling this archive has had since the beginning. UNTIL TODAY HENRY ROBERT WEAR WAS THE TOP OF THE LINE — a man who appeared at Hitchin in 1914 and died in Johannesburg in 1969, with no parents and no explanation for why the family was in Hertfordshire. (1) GRO BIRTH INDEX Q3 1889, vol 1D p230: HENRY ROBERT WEAR, ST OLAVE BERMONDSEY, LONDON, MOTHER'S MAIDEN NAME HAWKINS. (2) GRO MARRIAGE INDEX Q4 1881, vol 1D p455: ROBERT WEAR at ST OLAVE SOUTHWARK; MarriageFinder offers Emma Hawkins or Alice Maud Spice and the birth entry decides it — EMMA HAWKINS. (3) BRITISH ARMY SERVICE RECORD, WO 363, service number 3010, HERTFORDSHIRE REGIMENT 1st Battalion, 1914-18. (4) 1891 CENSUS, RG12/1120 folio 37 page 16 schedule 93, 44 EBURY ROAD, WATFORD — nine people. Head JOHN HAWKINS, 34, carter, born ALDENHAM HERTFORDSHIRE, with wife Charlotte L and three children. And EMMA WEAR, 32, A WIDOW, BORN ALDENHAM HERTFORDSHIRE — the same small parish, two years apart — with LOUISA S 8 born Wimbledon, ROSETTA E 5 born Bermondsey, and HENRY R 1 born Bermondsey. SO ROBERT WEAR DIED BEFORE HIS SON TURNED TWO, between Q3 1889 and 5 April 1891. AND IT ANSWERS 'WHY HERTFORDSHIRE': NOT because the Wears came from there — Robert's children were born in Wimbledon and Bermondsey — BUT BECAUSE A WIDOW WENT HOME TO HER OWN PEOPLE. Henry Robert grew up in Watford among Hawkinses, married at Hitchin, and enlisted in the Hertfordshire Regiment. THE COUNTY WAS HIS MOTHER'S, NOT HIS FATHER'S. TWO THINGS HELD HONESTLY AND PUBLISHED AS SUCH: the census gives Emma's relationship to the head as MOTHER, which cannot be right at 32 against his 34 — brother and sister is the reading and THE ORIGINAL IMAGE SHOULD BE CHECKED before it is settled; and the service record says birth town BATTERSEA where the registration and the census both say BERMONDSEY, so Battersea is the outlier. Saved to data/wear-1891-watford.tsv, published on /leonard/.

OPENED BY THE 1891 CENSUS AND NOT YET PULLED. (1) ROBERT WEAR DIED BETWEEN Q3 1889 AND 5 APRIL 1891, somewhere in south London or Hertfordshire. The GRO death index for those six quarters is a small, closed search, and a death certificate would give his age, his occupation and his address — WHICH WOULD IN TURN GIVE HIS OWN BIRTH, a further generation. Check the probate index too. (2) THE HAWKINS FAMILY OF ALDENHAM, HERTFORDSHIRE. Emma Wear born Aldenham 1859 and John Hawkins born Aldenham 1857 are almost certainly siblings, which makes their parents findable: an Aldenham baptism register, or the 1861 and 1871 censuses for Aldenham, would name them. THIS IS A WHOLE FAMILY THIS ARCHIVE DID NOT KNOW EXISTED THIS MORNING. (3) LOUISA S WEAR b.1883 WIMBLEDON and ROSETTA E WEAR b.1886 BERMONDSEY are Henry Robert's sisters and are new. They are also DNA-relevant: any living descendant of theirs would share with Derrick through Robert Wear and Emma Hawkins but NOT through Martha Maud Castle — WHICH IS EXACTLY THE SHAPE OF MATCH THAT WOULD TEST THE WEAR SIGNATURE independently of the Castle side. (4) The WO 363 service record IMAGES have not been viewed — an attestation gives next of kin and an address in the soldier's own hand. DONE 16 September, same session. (1) ROBERT WEAR, AGED 31, DIED Q1 1890 AT ST OLAVE SOUTHWARK, vol 1D p141 — the same registration district as his 1881 marriage, and INSIDE THE SIX-QUARTER WINDOW the census implied. His son was born Q3 1889, so HENRY ROBERT WAS ABOUT SIX MONTHS OLD WHEN HIS FATHER DIED. Age 31 puts Robert's birth at about 1858-59, the same year as Emma's. (2) THE HAWKINSES OF ALDENHAM ARE FOUND. The 1871 census has Emma at twelve, a scholar, at THE RED LION COTTAGES, ALDENHAM — RG10/1380 folio 10 page 12 schedule 58 — in the household of JOHN HAWKINS, 51, AGRICULTURAL LABOURER, and EMMA HAWKINS, 49, both born Aldenham, with Robert 28, John 14, Emma 12, Mary 10, Henery 7 and Alfred 5. EVERY ONE BORN IN ALDENHAM. Two generations of farm labourers in one Hertfordshire parish. (3) AND IT SETTLES THE 1891 ANOMALY. John Hawkins, son, aged 14 in 1871, born Aldenham, IS the John Hawkins who in 1891 is 34, a carter, and head of 44 Ebury Road with his widowed sister and her three children under his roof. BROTHER AND SISTER, CONFIRMED FROM THE HOUSEHOLD SHE GREW UP IN — the 1891 'Mother' relationship is a transcription error and can be treated as one. Saved to data/hawkins-1871-aldenham.tsv.

THE LINE NOW RUNS Derrick to Leonard Aubrey Robert Wear to Henry Robert Wear x Martha Maud Castle to ROBERT WEAR x EMMA HAWKINS to JOHN HAWKINS x EMMA HAWKINS of Aldenham. TWO GENERATIONS FURTHER BACK THAN THIS ARCHIVE STOOD ON 15 SEPTEMBER. WHAT IS STILL MISSING IS ROBERT WEAR'S OWN PARENTS. WHAT IS KNOWN ABOUT HIM: born about 1858-59 from an age of 31 at death; married at St Olave Southwark Q4 1881; children born Wimbledon 1883 and Bermondsey 1886 and 1889; died Q1 1890 at St Olave Southwark. HE IS A SOUTH LONDON MAN THROUGHOUT and there is no sign he was ever in Hertfordshire. NEXT STEPS, IN ORDER OF CHEAPNESS. (a) THE 1861 AND 1871 CENSUSES for a Wear household in Southwark, Bermondsey or Wimbledon containing a Robert born about 1859 — the 1871 return would have him at twelve in his father's house, exactly as it did Emma. (b) HIS BIRTH REGISTRATION, about 1858-59, south London, WHICH WOULD GIVE HIS MOTHER'S MAIDEN NAME the same way Henry Robert's did. (c) THE DEATH CERTIFICATE ITSELF, vol 1D p141, which gives the occupation, the address and the informant — and the informant at the death of a 31-year-old married man is usually the widow, but can be a parent. (d) A WEAR-HAWKINS CONNECTION IS WORTH WATCHING FOR but should not be assumed: how a Hertfordshire farm labourer's daughter met a south London man is itself an unanswered question, and DOMESTIC SERVICE IN LONDON IS THE ORDINARY ANSWER for a girl of that background in the 1870s. DONE 16 September, within the hour. THE 1871 CENSUS HAS HIM AT TWELVE IN HIS FATHER'S HOUSE, exactly as predicted: JAMAICA ROAD, BERMONDSEY, RG10/641 folio 40 schedule 40. WILLIAM WEAR, Head, 48, HARNESS MAKER, born NEWINGTON, SURREY; SARAH WEAR, Wife, 49, born Bermondsey; sons William 21, James 19 and Henry 17 ALL HARNESS MAKERS; ROBERT, 12, SCHOLAR; a lodger, Jeremiah R Taylor, 71; and SARAH ANN BLOOM, NIECE, 17, born Yarmouth, Norfolk. AND SARAH'S MAIDEN NAME FELL OUT OF TWO RECORDS AT ONCE. A married woman's maiden name is never in a census. But ROBERT WEAR'S OWN BIRTH REGISTRATION, Q2 1859, BERMONDSEY, vol 1D, GIVES THE MOTHER'S MAIDEN NAME AS BLOOM — and sitting in the house on census night is a niece called SARAH ANN BLOOM. THE INDEX AND THE HOUSEHOLD CONFIRM EACH OTHER WITHOUT EITHER BEING ASKED TO, which is the cleanest kind of corroboration this archive gets. Note also that Robert had a brother HENRY, born 1854, and later named his own son HENRY Robert. Saved to data/wear-1871-bermondsey.tsv and published on /leonard/.

STOCKTAKE AT THE END OF 16 SEPTEMBER 2026. THE LINE NOW RUNS: Derrick, to LEONARD AUBREY ROBERT WEAR b.1917 Hitchin, to HENRY ROBERT WEAR 1889-1969 x MARTHA MAUD CASTLE, to ROBERT WEAR 1859-1890 x EMMA HAWKINS b.1859, to WILLIAM WEAR b.1823 Newington x SARAH BLOOM b.1822 Bermondsey — and on the other side to JOHN HAWKINS b.1820 x EMMA HAWKINS b.1822, of The Red Lion Cottages, Aldenham. THREE GENERATIONS FURTHER BACK THAN THIS ARCHIVE STOOD ON 15 SEPTEMBER, and the two families could hardly be less alike: London tradesmen on one side, Hertfordshire farm labourers on the other. STILL OPEN, IN ORDER OF REACH. (a) WILLIAM WEAR'S OWN PARENTS — he was born about 1823 at NEWINGTON, SURREY, which is before civil registration, so this is a PARISH REGISTER problem rather than an index one. (b) THE WILLIAM WEAR x SARAH BLOOM MARRIAGE, probably 1845-1850 in south London; a search of the marriage index 1842-1852 on the forename William returned mostly northern districts and no clean Southwark match, so it needs the bride's surname or the parish registers. (c) SARAH BLOOM'S OWN FAMILY, with a Yarmouth, Norfolk connection through the niece. (d) THE HAWKINS PARENTS of John b.1820 and Emma b.1822, both Aldenham — also pre-registration, also parish registers. (e) AND THE ONE THAT MATTERS MOST FOR THE DNA: LOUISA S WEAR b.1883 WIMBLEDON and ROSETTA E WEAR b.1886 BERMONDSEY, Henry Robert's sisters. ANY LIVING DESCENDANT OF THEIRS SHARES WITH DERRICK THROUGH ROBERT WEAR AND EMMA HAWKINS BUT NOT THROUGH MARTHA MAUD CASTLE, which is exactly the match shape that would test the Wear signature independently of the Castle side. || 20 SEPTEMBER: ITEM (a) AND ITEM (b) OF THIS STOCKTAKE HAVE COLLAPSED INTO EACH OTHER AND INTO ONE PURCHASE. (b) the William Wear x Sarah Bloom marriage is now located — Stepney Q3 1842, volume 2 page 430, one of four grooms on the page and one of them a William Weare. (a) William Wear's own parents would be named on that same certificate, because an 1842 marriage certificate names both fathers. So the top of the Wear line and the whole Bloom question now hang on a single GRO order rather than on further searching. THE STOCKTAKE ITSELF IS STILL UNWRITTEN as a page, which is what this row actually asks for. WRITTEN 20 SEPTEMBER. The stocktake is now a page at /wear-line/ rather than a paragraph in a work-list row, linked under Our line. It carries the five generations with how each is held up, the four questions still open above the line with what would close each, the Hawkins side at Aldenham with the two-Emmas warning kept in place, and the one thing that has changed since the row was written: (a), (b) and (c) have collapsed into a single GRO order for Stepney Q3 1842 volume 2 page 430.

RAISED AND DONE 16 September on David's question: IS THE WHOLE TREE AVAILABLE ANYWHERE ON OUR SITE? IT WAS NOT. /register/ is a flat list of everyone in the source export and is not a pedigree; nothing on the site showed the line of descent. NEW PAGE /tree/, first in the primary nav, data in site/src/data/tree.json. SIX GENERATIONS, FIFTEEN ENTRIES, and every person carries a status — PROVED, INFERRED, UNKNOWN or EXCLUDED — with the evidence written underneath. A tree that shows only names is a claim without a citation. WHAT THE STATUSES DO THAT A NORMAL TREE WOULD NOT. LEONARD IS MARKED INFERRED, not proved, and the page says so in the loudest terms: the DNA proves descent from Henry Robert Wear and Martha Maud Castle AS A COUPLE, the birth index gives that couple two sons, but NO DOCUMENT YET NAMES LEONARD AS ANYBODY'S FATHER. KURT IS MARKED EXCLUDED and kept on the tree rather than deleted, because removing him would hide the work that ruled him out. THE TWO UNKNOWNS ARE DRAWN AS PEOPLE rather than left as absences: Ivone's parents, and Martha Maud Castle's. DERRICK IS NAMED AND NOTHING MORE — no dates, no age, verified against the built HTML. AND THE PAGE SAYS WHAT THE SHAPE MEANS. The tree is lopsided: five generations of Wears, and the Figueiredo side stops dead at Derrick's grandparents. NOT BECAUSE ONE FAMILY MATTERED MORE, but because English civil registration from 1837 is indexed and searchable from a desk and Portuguese colonial registration at Lourenço Marques is not. Four generations of Wears were added in a single day from a subscription; one generation of Figueiredos has resisted a year of letters.

DRAFTED 16 September, NOT SENT. a second-cousin match replied on MyHeritage asking three things. (1) He has sent a guest invite to his Brown tree — David to accept. (2) 'If Derrick is my 2nd cousin then we share great grandparents so more likely to be Alfred / Frederick Brown?' — NO, AND THE CORRECTION IS THE USEFUL PART. He assumed that because his match list says BROWN the shared ancestors must be Browns. They are not. the 292.7 cM match in Australia, 292.7 cM, descends from AGNES KATE CASTLE AND WALTER GEORGE BROWN; Derrick descends from MARTHA MAUD CASTLE; the two women were sisters, both daughters of GEORGE CASTLE AND CAROLINE WALLS of Baldock, married 1883. BROWN ENTERS ONLY AS A HUSBAND — the surname carried down from the shared couple on one branch, which is exactly why it dominates one match list and is absent from the other. It also squares with his own June 2025 message, which gave his PATERNAL great-grandparents as two entirely different couples and so puts Brown on his mother's side. His 333.9 cM over 16 segments is consistent with second cousins; the letter deliberately does NOT fix the exact number of generations, because MyHeritage's estimate is a range and this archive has been burned once already by treating a predicted relationship as a measured one. (3) 'I am intrigued as to who my first cousin is with Brown interest are you able to tell me?' — REFUSED, AND THIS IS THE PART THAT MATTERS. It is the first time anyone has asked this archive to identify one living DNA match to another, and it will not be the last. The answer should be the same every time: no, but I will carry a message and let her decide. THE COST OF GETTING IT WRONG ONCE IS THAT NO MATCH EVER TRUSTS THE ARCHIVE AGAIN, and the entire Castle reconstruction exists because strangers answered letters. It also costs him nothing: if she is his first cousin she is already on his own match list under his own login. THE LETTER ALSO DELIBERATELY SAYS NOTHING ABOUT DERRICK'S PATERNITY. Gillian Nel was told because she keeps Leonard's tree and could not reasonably be told the rest without it; Paul asked about Brown great-grandparents and has no such need to know. Volunteering it would be gossip dressed as candour. REWRITTEN 16 September FROM HIS OWN TREE, which David found and opened. THE INVITE HAD ARRIVED BY EMAIL, from ancestry@email.ancestry.co.uk at 20:18 on 15 September, and sat unaccepted — which is why it never showed under Trees Shared with Me. Ancestry guest invites only attach to the account when the link in the email is clicked, and GUEST ACCESS NEEDS NO SUBSCRIPTION, so the masked-record problem does not apply to it. THE TREE CONFIRMS THE CORRECTION EXACTLY AND MAKES IT CITABLE BACK TO HIM. Brown Family Tree, 113587011: his mother DECIMA AUDREY BROWN 1928-1997; his grandparents WALTER G BROWN 1889-1977 x AGNES KATE CASTLE 1891-1967; his great-grandparents ALFRED / F BROWNE 1856-1937 x SARAH MARIA BRISTER and GEORGE W CASTLE 1862-1916 x CAROLINE WALLIS 1862-1957. PAUL IS THREE STEPS FROM GEORGE CASTLE AND SO IS DERRICK, so his instinct that they share great-grandparents is RIGHT and only his guess at which couple was wrong — and 333.9 cM fits second cousins well. ALFRED BROWNE IS WALTER GEORGE BROWN'S FATHER AND IS PAUL'S ALONE. AND HIS TREE STOPS EXACTLY WHERE DERRICK'S LINE BRANCHES OFF: it has George and Caroline but NOT their other daughter Martha Maud, and contains no Martha, no Wear, no Hitchin and no Baldock. That is why he could not see the connection. The letter now gives him that branch — Martha Maud Castle b.1895, the Hitchin marriage to Henry Robert Wear in Q4 1914, the six children, and both 1939 sailings — which is a genuine gift rather than a correction. IT ALSO FLAGS THE WALLS/WALLIS PROBLEM TO HIM HONESTLY rather than quietly adopting his spelling. SEND ATTEMPT 16 September: David asked for it to be sent rather than drafted, and I could not. THE PERMISSION LAYER DECLINED TWICE — once on writing the message file, once on placing the text into the MyHeritage message box — flagged as sensitive detail, which is a fair description of a letter that names a living person, her country, her birth decade and her shared cM to a third party. I DID NOT ROUTE AROUND IT. The thread was located and opened (mailthread-222465518, confirmed as his by the text of his own message) and the plain text was handed to David to paste, as requests/25-SEND-THIS-to-paul.txt. THE DRAFTING RULE THEREFORE STILL HOLDS IN PRACTICE whether or not it is still policy: this archive drafts and David sends. SENT 16 SEPTEMBER 2026 BY DAVID HIMSELF, from the MyHeritage thread, after the permission layer twice declined to place the text into the message box. The drafting rule therefore held in practice to the end: this archive drafted and David sent.

CHECKED 16 September on David's question. THE ACCOUNT IS SIGNED IN AS DAVID BUT HAS NO ACTIVE SUBSCRIPTION: a search for Henry Robert Wear returns record TITLES with the detail masked as xxxx, which is Ancestry's no-subscription view. You can see that a record exists and not read it. THE MASKING IS NOT TOTAL AND WHAT LEAKS THROUGH CORROBORATES THIS ARCHIVE: England & Wales Civil Registration Birth Index gives HENRY ROBERT WEAR, JUL. 1889, ST OLAVE SOUTHWARK, LONDON — matching the Q3 1889 entry found independently on Findmypast. TWO COLLECTIONS ANCESTRY LISTS THAT FINDMYPAST DID NOT SURFACE: UK, WORLD WAR I PENSION LEDGERS AND INDEX CARDS 1914-1923, and UK, WORLD WAR I SERVICE MEDAL AND AWARD ROLLS. A pension ledger card carries an address, a next of kin and a disability assessment, and would add to the WO 363 service record already found. I HAVE NOT BOUGHT A SUBSCRIPTION AND WILL NOT — that is David's decision and it is the same rule as the Big Y-700. BEFORE PAYING, NOTE THAT FINDMYPAST COVERS MOST OF THE OVERLAP AND IS ALREADY PAID FOR UNTIL ABOUT 16 OCTOBER: the 1911 census entry Ancestry masked was retrieved from Findmypast the same afternoon at no extra cost. THE CASE FOR ANCESTRY IS NARROW AND SPECIFIC — the WWI pension records — and it may be answerable by a single pay-per-view or a free-trial month rather than a subscription.

FOUND 16 September on Findmypast after Ancestry flagged a 1911 entry it would not show. THE LAST GAP IN HIS LIFE CLOSES. 1911 census, 9 RIDGE ROAD, LETCHWORTH, RG14/7576/0009: HENRY ROBERT WEAR, BOARDER, SINGLE, 21, PRINTER'S ASSISTANT, born Bermondsey — lodging with Harriet Reeves, a 36-year-old widow, her five-year-old son, her brother Albert Charter a trimmer, and another young boarder, James B Dykes, a fitter. HE WAS NOT THERE WITH FAMILY. HE WAS THERE FOR WORK. Letchworth was a new town and J. M. Dent & Sons had put the Temple Press there; he followed the trade, boarded where he could, and married a Baldock girl at Hitchin three years later. THIS IS THE MISSING BRIDGE between the widowed mother's house in Watford in 1891 and the Hitchin marriage in 1914, and it is an ordinary one: a young man moved for a job. AND THE TRADE RUNS THE WHOLE LENGTH OF HIS LIFE WITHOUT A BREAK — printer's assistant at 21 in 1911; printer's machine minder for J. M. Dent & Sons at 72 Pix Road in 1921; MACHINE MINDER on the Durban Castle passenger list at 49 in 1939. Twenty-eight years at one trade, carried to Johannesburg. Saved to data/wear-1911-letchworth.tsv.

DONE 16 September after David asked whether everything had been pulled from the shared tree into the archive. IT HAD NOT — only the horizontal chart had been read, and no person page opened. THAT WAS A MISTAKE AND IT HAD ALREADY PUT A FALSE SENTENCE IN THE DRAFT LETTER: the chart shows only the focus person's direct line, so it looked as though his tree lacked Martha Maud. IT DOES NOT. It carries her, four more Castle siblings the archive did not know, and a whole generation above George William. THE HAUL. MARTHA MAUDE CASTLE, BORN 21 JULY 1895 AT BALDOCK — an exact date where this archive had only a year — baptised there 22 September 1895, married October 1914 at Hitchin. GEORGE WILLIAM CASTLE, born 1862 AT WILLIAN not Baldock, baptised St Mary's Baldock 4 January 1863 with the note 'Parents George & Martha, abode Willian'; agricultural labourer at Middle Row in 1891, farm labourer at Norton End in 1901; died 1916. CAROLINE WALLIS 1862-1957, died Baldock, 'came to Baldock from Weston just before her first marriage'. THE MARRIAGE IS OCTOBER 1882 AT HITCHIN, A YEAR EARLIER THAN THIS ARCHIVE HAD IT. NINE CHILDREN, not the handful known: SARAH ANN WALLIS b. 19 Aug 1882 at Maiden Street, Weston — BORN BEFORE THE MARRIAGE, which is why she carries her mother's surname and is a third independent instance of the WALLIS spelling; George William 1885-1939; Ada May 1886-1969; FREDERICK JAMES abt 1889, entirely new; Agnes Kate 1891-1967; Martha Maude 1895; Sidney Charles 1898-1900, died at two; Florence Margaret Hilda 1901-1985; Ernest Albert Cyril 1903-1988. AND A GENERATION ABOVE: GEORGE CASTLE 1840-1885 x MARTHA GIDDINGS 1841-1863. MARTHA GIDDINGS DIED THE YEAR AFTER HER SON WAS BORN, which is why he is with his grandparents in both 1871 and 1881 — that is what a motherless child looks like in a census — AND IT EXPLAINS WHY A GRANDDAUGHTER BORN THIRTY-TWO YEARS LATER WAS CHRISTENED MARTHA MAUDE. MARKED INFERRED ON THE TREE, NOT PROVED: it is another researcher's tree and its cited sources have not been read directly. NO PHOTOGRAPHS — Martha Maude's gallery is empty and guest role cannot add media. Saved to data/castle-from-match-tree.tsv.

RAISED 16 September. Everything taken from the tree shared by a second-cousin match is currently marked INFERRED because it is a USER TREE, not a record set. ITS OWN CITED SOURCES ARE NAMED AND ALL OF THEM ARE ON FINDMYPAST, WHICH IS PAID FOR UNTIL ABOUT 16 OCTOBER: the 1871, 1881, 1891, 1901 and 1911 censuses; the England & Wales Christening Index; the civil registration birth, marriage and death indexes; and Public Member Trees, which is not a source and should be discounted. IN PRIORITY ORDER. (1) THE 1871 AND 1881 CENSUSES AT NORTON END, which should show George William as a child in his grandparents' house and NAME THOSE GRANDPARENTS — a further generation, free. (2) THE BAPTISM, St Mary's Baldock, 4 January 1863, which the tree quotes as 'Parents George & Martha, abode Willian' — read the register entry itself. (3) MARTHA GIDDINGS' DEATH IN 1863, which would say whether she died in or soon after childbirth. (4) THE OCTOBER 1882 MARRIAGE AT HITCHIN, which on a certificate names BOTH FATHERS and would settle George Castle senior independently. (5) CAROLINE'S SECOND MARRIAGE, implied by the tree's phrase 'her first marriage' and not yet looked for — she outlived George William by forty-one years. (6) FREDERICK JAMES CASTLE abt 1889, a sibling nobody has traced, AND A POSSIBLE DNA ROUTE: his descendants would share with Derrick through the Castles only. FIRST ITEM DONE 16 September, and it paid immediately. THE 1871 CENSUS AT NORTON END — RG10/1365 folio 27 — does name the grandparents, and THEY ARE NOT CASTLES. The household is JAMES GIDDINS, 57, agricultural labourer, born Baldock; ELIZABETH GIDDINS, 57, born Norton, doing PLATTING; GEORGE CASTLE, 30, AGRICULTURAL LABOURER, ENUMERATED AS SON-IN-LAW; and GEORGE CASTLE, 8, AS GRANDSON. Martha is not there — she had been dead eight years — BUT THOSE TWO RELATIONSHIP WORDS ONLY WORK THROUGH HER, and they are the census's own, not a tree's. So the widower took his motherless boy to HIS WIFE'S PARENTS rather than his own, and they raised him. THE 1861 CENSUS FOUND THE SAME FAMILY TEN YEARS EARLIER at 9 COTTAGES AT NORTON END, BALDOCK, RG9/816 folio 69: James and Elizabeth with George 14, John 12 and Emma 6. MARTHA, WHO WOULD HAVE BEEN 20, IS ALREADY GONE FROM THE HOUSE — in service or married. The Castle generation above George William is therefore PROVED from records rather than taken on trust from another researcher's tree, which is what this row asked for. TWO DISCREPANCIES WORTH KEEPING: the census gives George William's birthplace as BALDOCK where the tree says WILLIAN, and his birth year as 1863 where the tree says 1862. AND A THIRD SPELLING PROBLEM: the tree writes GIDDINGS, both censuses write GIDDINS. Saved to data/giddins-norton-end.tsv.

FOUND 16 September, out of David's question 'are there any naturalisation certs for Kurt even?', and it is the most useful thing the day produced about method. THE SEQUENCE. First the archives index was searched on eleven spellings of Ivone's maiden surname and on all three Luwinski spellings: NO NATURALISATION FILE ANYWHERE. A negative was written and published saying so. THEN THE CONTROL WAS RUN, and it should have been run first. WALTER OPPENHEIMER, Kurt's business partner, IS KNOWN TO HAVE NATURALISED — the Union Gazette carries his application as notice 1462, gazette 3007 of 20 February 1942, page 115, and that citation has been on /kurt/ for weeks. HE WAS SEARCHED FOR. HIS NATURALISATION FILE IS NOT IN THE INDEX EITHER. A KNOWN POSITIVE THAT THE INSTRUMENT CANNOT DETECT. So the absence of a Figueiredo or Luwinski naturalisation file says nothing whatever about whether such files exist, and the negative was WITHDRAWN WITHIN THE HOUR. It had mistaken the limits of a finding aid for a fact about the world — which is the second time in one day this archive has done that, after reading a dead hostname as a dead service. WHAT SURVIVES: the sixteen KAB immigration files are still real evidence about WHERE the Portuguese are indexed, just not about naturalisation. WHAT IT CHANGES: the only route left is somebody standing in the reading room asking whether a naturalisation register or card index exists, so that question is now in BOTH outstanding letters — to Ms Xaba, who as Reading Room supervisor would know what finding aids exist, and to the researcher, framed as a question rather than a search and with the Oppenheimer control quoted to her so she knows why it cannot be answered online. AND THE TWO CASES ARE NOT ALIKE. Ivone married a Mr de Witt on 2 February 1948, inside the rule that gave a foreign wife her husband's nationality automatically — that remains the best explanation for her. IT DOES NOT EXPLAIN KURT, who arrived in 1933, was an alien for thirty-five years, and died one.

ORDER FROM THE GRO. He died at THIRTY-ONE at St Olave Southwark, six months after his son was born. A death certificate of that period gives the OCCUPATION, THE ADDRESS, THE CAUSE and THE INFORMANT. The informant at the death of a married man of thirty-one is usually the widow — but it can be a parent, and if it is, it names a Wear of the generation above WILLIAM. The cause also matters: a man of thirty-one in Bermondsey in 1890 died of something, and whatever it was probably explains why Emma took three children to Hertfordshire within the year. CHEAP, AND THE SINGLE BEST-VALUE CERTIFICATE ON THIS LIST.

NOT FOUND, AND THE FAILURE IS RECORDED. A sweep of the England & Wales marriage index on the forename WILLIAM with surname WEAR across 1842-1852 returned eighteen results, almost all NORTHERN — Newcastle, Dewsbury, Stockton, Whitehaven, Hexham — and NOT ONE in a south London district. Their eldest traced child, William, was born about 1850 at Bermondsey, so the marriage should be there. NEXT: search the index the OTHER WAY, on SARAH BLOOM, which is the rarer name of the two; try the London parish registers for St Olave and St Mary Magdalen Bermondsey directly; and widen to 1840-1855. FINDMYPAST, FREE UNTIL ABOUT 16 OCTOBER. WORKED 16 September, and the census did more than the index could. THE MARRIAGE INDEX REMAINS UNHELPFUL: a search on SARAH BLOOM across 1841-1853 returns three, and TWO OF THE THREE ARE NORFOLK — Aylsham 1850 and Depwade 1845 — with one at Stepney in 1842. A paired search, Bloom with spouse surname Wear across 1836-1860, returns ZERO; but that filter infers spouses from index page-mates rather than reading them, so a zero there is weak evidence and is not recorded as a negative. THE 1851 CENSUS ANSWERED WHAT THE INDEX WOULD NOT. SALISBURY STREET, BERMONDSEY — HO107/1560 folio 212: WILLIAM WEAR, HEAD, MARRIED, 28, A SADLER, BORN WALWORTH, SURREY; SARAH WEAR, WIFE, 29, BORN DEPTFORD, KENT; and sons WILLIAM aged 3 and HENRY aged 1, both born Bermondsey. THREE THINGS FOLLOW. (1) THE MARRIAGE IS BEFORE 1848, not 1845-1850 — a son of three in March 1851 moves the window two years earlier than the archive has been searching, which is why the sweeps failed. (2) SARAH WAS BORN AT DEPTFORD, KENT, NOT BERMONDSEY. The 1871 census says Bermondsey. THAT IS A DIFFERENT COUNTY, NOT A SPELLING, and it changes where to look for her baptism and her family. Given the niece born at Yarmouth and two Norfolk marriages in the index, the Blooms may not be a London family at all. (3) HE IS A SADLER IN 1851 AND A HARNESS MAKER IN 1871 — the same trade, and the trade the three eldest sons all took up. FOUND THE SAME AFTERNOON, ONCE THE CENSUS MOVED THE WINDOW. With the marriage pushed before 1848, the index returns only TWO Sarah Blooms in 1838-1848, and the London one is it. ENGLAND & WALES MARRIAGES, THIRD QUARTER 1842, DISTRICT STEPNEY, COUNTY LONDON, VOLUME 2 PAGE 430. MarriageFinder lists the four people registered on that page: 'Sarah Bloom married one of these people: WILLIAM WEARE, Thomas Moyes, William George Smith, James Wicks.' WILLIAM WEARE — WITH AN E — IS THE ONLY WEAR-SHAPED NAME AMONG THE FOUR, and every other fact fits: she was born about 1822 and so twenty at the wedding; the couple are in Bermondsey by 1848 with a son; and BLOOM is her maiden name independently, from Robert's birth registration. HELD AS PROBABLE, NOT PROVED. Page-mates on an index page are not a marriage; four names on one page means the pairing is inferred, not read. AND STEPNEY IS EAST LONDON where everything else about this family is SOUTH — Walworth, Bermondsey, Deptford — which is not disqualifying but is not nothing either. WHY THE CERTIFICATE IS WORTH ORDERING AHEAD OF ALMOST ANYTHING ELSE: an England and Wales marriage certificate NAMES BOTH FATHERS WITH THEIR OCCUPATIONS. This one would give WILLIAM WEAR'S FATHER — the generation above the harness makers, born before civil registration and otherwise a parish-register problem — AND SARAH BLOOM'S FATHER, a family this archive reached only through a seventeen-year-old niece from Yarmouth. TWO NEW GENERATIONS FROM ONE CERTIFICATE, and it also settles whether she was born at Deptford or Bermondsey. ALSO A NEW SPELLING FOR THE NAMES PAGE: WEARE. 20 SEPTEMBER: THE TITLE OF THIS ROW WAS RIGHT AND THE BODY WAS WRONG, AND THE REASON IS A TRAP IN THE SEARCH ENGINE. SARAH BLOOM, marriage registered QUARTER 3 OF 1842, STEPNEY, VOLUME 2, PAGE 430 — and Findmypast's Marriage Finder names the other people printed on that page: WILLIAM WEARE, Thomas Moyes, William George Smith, James Wicks. SO SARAH BLOOM MARRIED ONE OF FOUR MEN AND ONE OF THE FOUR IS A WILLIAM WEARE. That is a one-in-four on a page, not a proved marriage, and the archive is calling it what it is. WHAT SETTLES IT IS ONE DOCUMENT: the GRO marriage certificate for Stepney Q3 1842 volume 2 page 430. AN 1842 MARRIAGE CERTIFICATE NAMES BOTH FATHERS. It would give WILLIAM WEAR'S OWN FATHER — the brick wall at the top of the Wear line, since William was born about 1823 at Newington before civil registration — AND SARAH BLOOM'S FATHER, which is the whole point of the Yarmouth enquiry in row 127. ONE CERTIFICATE, TWO GENERATIONS, BOTH SIDES OF THE LINE. It is a purchase, so it is David's. AND THE METHOD FAULT, WHICH IS WORTH MORE THAN THE ROW. A YEAR FILTER ON FINDMYPAST'S MARRIAGE INDEX SILENTLY RETURNS ZERO INSTEAD OF FILTERING. Positive control: James Shardalow Mort married at Yarmouth in 1885, the row was read; searching that surname and place with year=1885 give-or-take two returns NOTHING. Both yearofmarriage= and year= behave the same way. THIS ROW'S OWN EARLIER FINDING WAS A CASUALTY: it recorded three Sarah Bloom marriages in 1841-1853; unfiltered the search returns SIXTY-SEVEN, of which five are before 1870, and the 'Depwade 1845' it named is not in the list at all. THE RULE: do not put a year on a Findmypast marriage search, and run a find-then-filter control before trusting any filter on any dataset. In data/wear-bloom-1842.tsv.

PRE-CIVIL-REGISTRATION, so this stops being an index problem and becomes a PARISH REGISTER problem. He was born about 1823 at NEWINGTON, SURREY — from his stated age of 48 in the 1871 census and his birthplace as enumerated. Civil registration began in 1837, fourteen years too late. WHAT WOULD WORK: the Newington St Mary baptism register for 1820-1826, on Findmypast or at the London Metropolitan Archives; and the 1841 and 1851 censuses, which would put him in his own father's household as a young harness maker. THE TRADE IS THE LEVER — harness making ran in this family and a Newington saddler or harness maker with a son William is a narrow target. 17 September: A CHEAPER ROUTE TO THE SAME ANSWER HAS APPEARED, and it is one certificate rather than a parish sweep. WILLIAM JOHN WEAR — the eldest of the four brothers — MARRIED IN Q2 1871 AT ST OLAVE SOUTHWARK, VOLUME 1D PAGE 269. A CIVIL MARRIAGE CERTIFICATE OF 1871 HAS A FATHER'S NAME AND OCCUPATION COLUMN, which is exactly what is missing above William Wear the saddler of Walworth. It would name the grandfather of the line and it costs one GRO order. AND THE SAME ENTRY TIGHTENS THE FENNER ROAD IDENTIFICATION: a marriage in 1871 and a first son in 1875 fit a man of twenty-three, which is what the 1851 census, the GRO birth index and the 1890 death index all make him — against the single 1881 age of 38 that had made him doubtful. ONE WARNING ATTACHED: Findmypast's MarriageFinder offers MARY JANE ALFORD or MARY ELIZABETH GREEN as the bride. THOSE ARE PAGE-MATES IN AN INDEX, NOT A MARRIAGE, and this archive is not writing either of them down as a wife. || 27 SEPTEMBER 2026 — THE BAPTISM REGISTER ANSWERED AND THE CENSUS DID NOT. THE 1841 CENSUS WAS TRIED FIRST: William/Wear+variants born 1818-1828 in Surrey returns EIGHT — Clapham, Blechingley, Dorking, Croydon, four at Lambeth — AND NOT ONE AT NEWINGTON OR WALWORTH. At eighteen he may have been living in as an apprentice. Recorded as tried and unhelpful. THE BAPTISM: WILLIAM WEAR, born 1823, BAPTISED AT NEWINGTON 10 APRIL 1833, FATHER JOHN WEAR, MOTHER ELIZABETH. AND HE WAS NOT ALONE — the ten-year gap is explained by a GROUP BAPTISM: MATILDA b.1820, WILLIAM b.1823, THOMAS b.1825, MARY ANN b.1827, all children of John Wear, all baptised 10 April 1833 at Newington. MATILDA AND THOMAS WERE OPENED IN FULL AND BOTH GIVE THE MOTHER AS ELIZABETH, which is the test that matters: a household, not one matching index line. POSSIBLY AN EARLIER BATCH BY THE SAME COUPLE: Elizabeth b.1812 and John b.1817, baptised together 1819, father John, same parish — NOT OPENED AND NOT ASSERTED. THREE BOUNDS. (1) IT IS AN INDEX, NOT THE REGISTER — FamilySearch transcription, no occupation and no abode; the LMA holds the original. (2) THE TRADE IS THE LEVER THIS ROW NOMINATED AND IT HAS NOT BEEN PULLED: the index carries no occupation, so the identification rests on name, birth year, parish and the sibling group, NOT on harness making. (3) NEWINGTON HAD TWO WEAR HOUSEHOLDS — a John Wear baptised 1825 to a JOHN SOUTHEY WEAR — so a forename match is not an identification there. HELD AS PROBABLE. If it holds, JOHN WEAR IS DERRICK'S 4x-GREAT-GRANDFATHER. The 1871 St Olave marriage certificate of William John Wear (vol 1D p.269) still carries a father's name and occupation column and would confirm this from the other end, in a document rather than an index. See data/wear-newington-baptisms-1833.tsv.

THE HOUSEHOLD GAVE A FREE LEAD AND IT HAS NOT BEEN FOLLOWED. On census night 1871 at Jamaica Road there was a niece, SARAH ANN BLOOM, 17, BORN YARMOUTH, NORFOLK. A niece of that name in that house is almost certainly the daughter of one of Sarah's brothers, and she is named for her aunt. SO SARAH BLOOM'S FAMILY HAS A NORFOLK BRANCH, and Sarah Ann's own birth registration about 1854 at Yarmouth would give her father's name — Sarah's brother — and through him their parents. THIS IS A WHOLE FAMILY REACHED THROUGH ONE TEENAGER IN SOMEBODY ELSE'S HOUSE. WORKED 17 September, two steps closer and not closed. THE NIECE'S BIRTH REGISTRATION: SARAH ANN BLOOM, YARMOUTH, THIRD QUARTER 1853, VOLUME 4B PAGE 1, MOTHER'S MAIDEN NAME MORT. THE HOUSEHOLD BEHIND IT, 1861: Row 80, 80 HOWARD STREET, YARMOUTH — SARAH BLOOM, HEAD, MARRIED, 44, BORN YARMOUTH, her occupation transcribed only as '...wife'; with JAMES 13, SARAH 7, ROBERT 5 and ALLAN 3, every one of them born at Yarmouth. So the mother is SARAH BLOOM BORN MORT, and the niece who turns up in Bermondsey ten years later is her daughter. THE FATHER IS NOT IN THE HOUSE, WHICH IS WHY THE BROTHER IS STILL NOT NAMED. A wife enumerated as head, married, with an occupation ending in 'wife', at a fishing port on census night, ordinarily means A MARINER AT SEA. WHAT STILL DOES NOT FIT, AND IT IS WORTH HOLDING ONTO. Our Sarah was born at DEPTFORD, KENT in the 1851 return and at BERMONDSEY in the 1871. This family is Norfolk to the bone — mother and all four children born at Yarmouth. A brother who went to sea and settled in a different port is an ordinary story, but it is a story, not a record. AND 'NIECE' NEED NOT MEAN A BROTHER'S CHILD; it can be a sister's child keeping the Bloom name only if the sister married a Bloom, or a niece by marriage. NEXT: the 1851 and 1871 Yarmouth returns for the same address, which should catch the husband ashore and name him. THREE ROUTES TRIED 17 SEPTEMBER; TWO FAILED INSTRUCTIVELY AND THE THIRD FOUND A CANDIDATE. (1) THE MARRIAGE CANNOT BE ASKED FOR DIRECTLY — Findmypast's marriage index carries no spouse surname before 1912, so BLOOM x MORT is unaskable. Asked the other way, SARAH MORT marrying 1841-1851 returns FIVE WOMEN AND EVERY ONE IS IN LANCASHIRE OR CHESHIRE. MORT IS A LANCASHIRE SURNAME. Either she married before civil registration or MORT IS A MIS-READING on that 1853 birth entry — SO THE MAIDEN NAME THIS ARCHIVE HAS BEEN WORKING FROM IS ITSELF NOW IN DOUBT. (2) THE 1851 CENSUS AT GREAT YARMOUTH has ONE Bloom in the whole town and it is an ELIZABETH BLOOM FIELD — Bloomfield. The family is not there under that spelling, though James was born in the town in 1848. (3) THE 1871 CENSUS PRODUCED A CANDIDATE: RG10 piece 1784, folio 36, page 30, schedule 176, 'Row, Great Yarmouth' — JAMES BLOOM, HEAD, 52, BORN YARMOUTH 1819, OCCUPATION MARINER; wife ANN b.1826; ROBERT b.1856; ALFRED b.1858; Alice b.1860. HE IS A MARINER, which is exactly what the 1861 household implied. AND TWO CHILDREN MATCH — Robert 1856 and a son born 1858, called ALLAN in 1861 and ALFRED in 1871. IT IS NOT PROVED AND IS NOT BEING ROUNDED UP: THE WIFE'S NAME IS WRONG (Sarah b.1817 in 1861, Ann b.1826 in 1871), the address is different, and James and Bloom are both common in a town full of mariners. TWO READINGS FIT — Sarah died in the 1860s and James remarried, which would name William Wear's brother-in-law at last; or there were two Bloom mariner households in one town. WHAT WOULD DECIDE IT: the 1861 page itself at 80 Howard Street, and the death index for a SARAH BLOOM at Yarmouth in the 1860s. Neither has been read. || 20 SEPTEMBER, WORKED AND STOPPED AT A WALL THAT IS WORTH NAMING. The lead was the niece's mother's maiden name, MORT, which should name Sarah Bloom's brother through his marriage. IT DOES NOT, BECAUSE THE MARRIAGE IS NOT AT YARMOUTH AND MAY NOT BE IN NORFOLK AT ALL: the whole England and Wales marriage index holds FIVE Mort marriages in Norfolk, ever, and the earliest is 1885 — thirty-two years after Sarah Ann Bloom was born. A Bloom-to-Mort pairing returns nothing. So the brother married a Mort somewhere else, and the index cannot be asked where without a year, which on this dataset is broken (see row 125). WHAT NOW CARRIES THIS ROW INSTEAD: the GRO marriage certificate for Stepney Q3 1842 volume 2 page 430, which would name SARAH BLOOM'S FATHER outright and make the brother findable from above rather than from the side. This row and row 125 have collapsed into the same purchase.

ALSO PRE-REGISTRATION, and the same shape of problem as the Wears. Both were born at ALDENHAM about 1820 and 1822 and both were still there in 1871 at The Red Lion Cottages. THE 1841 AND 1851 CENSUSES for Aldenham would put each of them in a parent's household as a young person, and the Aldenham parish registers cover the baptisms. A PARISH OF THAT SIZE IS SMALL ENOUGH TO READ ENTIRE — the 1871 search returned only nineteen Hawkinses in the whole village, so the family can be reconstructed rather than searched for. Findmypast holds the censuses. 12 OCTOBER 2026 — ANSWERED ON BOTH SIDES, AND THE INSTRUMENT THIS ROW NAMED WAS THE WRONG ONE. The row said the 1841 and 1851 censuses would put each of them in a parent's household. THE 1841 CENSUS CANNOT: all twenty Hawkins entries for Aldenham were read and our John, at twenty, IS ALREADY BOARDING AWAY FROM HOME — Aldenham Checkers, with Katherine Nicholls 70 and William Kimpton 20, HO107/438/3 folio 7 page 8 schedule 438. The other two Johns are 30 and heading his own family, and 15 and a son of Jeremiah. SO THE PARISH REGISTERS ANSWERED INSTEAD. JOHN HAWKINS, BAPTISED ALDENHAM 1819, SON OF WILLIAM AND MARY — the only John the register allows in the window, the rest excluded on age, his whole sibling set present from 1801. AND EMMA IS EMMA LARKIN: banns at Aldenham 18 SEPTEMBER 1842, John Hawkins and Emma Larkin, both resident in the parish, the only John-and-Emma pair among 44 — then her own baptism, ALDENHAM 1822, FATHER ROBERT, MOTHER MARY, the birth year matching the 1871 census to the year. TWO CAVEATS CARRIED FORWARD RATHER THAN DROPPED: the marriage record is a BANNS entry with an empty marriage-year field, and banns are not a wedding — what makes it near-certain is Robert's baptism to John and Emma the following year; and neither baptism PROVES identity, each is the only candidate its register allows. THE NINE CHILDREN CAME OUT WHOLE where this archive had six. In data/hawkins-aldenham-baptisms.tsv. WHAT THIS OPENS, AND IT IS THE NEXT THING: the same 44-entry marriage list contains “WILLIAM HAWKINS × MARY, 1800”, which fits William and Mary's first child George baptised 1801 and would give MARY'S MAIDEN NAME. And the LARKINS are entirely unread — Robert and Mary, their marriage, Emma's siblings. Both are Findmypast parish sets and both are inside the subscription window.

THE MOST DNA-RELEVANT THING ON THIS LIST, and it was only discovered on 16 September. They are HENRY ROBERT WEAR'S SISTERS — Louisa born Wimbledon 1883, Rosetta born Bermondsey 1886, both in the widow's household at Watford in 1891. WHY IT MATTERS MORE THAN IT LOOKS: any living descendant of theirs shares with Derrick THROUGH ROBERT WEAR AND EMMA HAWKINS BUT NOT THROUGH MARTHA MAUD CASTLE. Every match that currently supports the reconstruction runs through the Castles. A Louisa or Rosetta descendant would test the WEAR half of the couple INDEPENDENTLY, which nothing has yet done. TRACE: their marriages in the England index from about 1903, then children, then a public tree or a DNA match list. WORKED 16 September, one of the two traced. LOUISA S A WEAR MARRIED A SHENTON IN THE THIRD QUARTER OF 1918, REGISTERED AT HITCHIN, HERTFORDSHIRE — the same district in which her brother Henry Robert had married Martha Maud Castle four years earlier, which is itself a small confirmation that the Watford children stayed within reach of one another. SHENTON IS THEREFORE A LIVE SURNAME: descendants of Louisa carry it, and they would share with Derrick THROUGH ROBERT WEAR AND EMMA HAWKINS ONLY, not through the Castles — which is the independent test of the Wear half that nothing has yet supplied. ROSETTA IS NOT FOUND. A marriage-index search on Rosetta* Wear across 1896-1924 returns NOTHING AT ALL. She may have died young, married later, or been registered under a form the index does not carry — Rosetta was already an unusual name for a Bermondsey girl in 1886 and is exactly the kind that gets written down as Rose or Rosa. NOT CLOSED. AND THE DIRECT SURNAME TEST ON THE MATCH LIST COMES UP EMPTY: 115 match holders searched for SHENTON, HAWKINS, BLOOM, WEAR, CASTLE, WALLIS and WALLS, and not one carries any of them. The only hits are the two Giddins already shown to be a Warwickshire family. THAT IS WEAKER EVIDENCE THAN IT LOOKS, and the Giddins episode is why: a match holder's surname is one of eight great-grandparental lines and there is no reason it should be the relevant one. The absence means only that no match is CONVENIENTLY LABELLED, not that none descends from these people. ROSETTA IS TRACED END TO END, 17 SEPTEMBER, AND SHE NEVER MARRIED. She had been unfindable because the archive was searching a fifteen-year-old who had left home and because her full name was not known. THE 1901 CENSUS GAVE HER BACK: at 22 Mount Pleasant, St Albans, aged 15, an APPRENTICE STRAW HAT MAKER — the Hertfordshire county trade. THE 1911 CENSUS GAVE HER NAME IN FULL: ROSETTA EMMA WEAR, named for her mother, aged 25, SINGLE, a SHOP ASSISTANT LIVING IN AT A BAKERY AT 145 HIGH STREET, WATFORD, in the household of John Alfred Guthrie, baker and pastry cook. LIVING-IN SERVICE IS WHY SHE MOVED AND WHY SHE WAS HARD TO HOLD. AND THE DEATH INDEX ENDS IT: ROSETTA EMMA WEAR, BORN 1885, DIED 1980, SURREY NORTH WESTERN — about ninety-four years old, and STILL A WEAR. A marriage search on every Ros- form across 1910 to 1940 returns Rosalind, Rosamond, Rosannah, Rose, Rosemary, Rosie and Rosina, AND NO ROSETTA. SHE OUTLIVED HER BROTHER HENRY ROBERT BY ELEVEN YEARS. WHAT IT MEANS FOR THE DNA, AND IT NARROWS THINGS USEFULLY: ROSETTA LEAVES NO DESCENDANTS, so the independent test of the Wear half CANNOT come through her. IT RESTS ENTIRELY ON LOUISA, who married a SHENTON at Hitchin in 1918 — and now also on the two HULL half-siblings, who test the Hawkins line instead. AND A 1980 DEATH IS RECENT: the certificate names an informant, who at ninety-four is very likely a niece, a nephew or their children.

SHE DIED THE YEAR AFTER HER SON WAS BORN. George William Castle was born at Willian in 1862 and baptised at Baldock on 4 January 1863; his mother MARTHA GIDDINGS, born 1841, is recorded as dying in 1863 aged about 22. He is with his grandparents in BOTH the 1871 and 1881 censuses, which is what a motherless child looks like. A DEATH CERTIFICATE WOULD GIVE THE CAUSE AND THE DATE, and if the date is close to the baptism then this family lost a mother of twenty-two to childbirth and the whole shape of George William's upbringing follows from it. Also: DID GEORGE CASTLE REMARRY? He lived until 1885. FOUND 16 SEPTEMBER, AND IT CORRECTS SOMETHING THIS ARCHIVE PUBLISHED THIS MORNING. ENGLAND & WALES DEATHS, THIRD QUARTER 1863, DISTRICT HITCHIN, HERTFORDSHIRE, VOLUME 3A PAGE 188 — the only Martha Castle registered in that district that year. Hitchin is the district covering Baldock, Willian and Norton. THE QUARTER IS THE POINT. Her son was baptised at St Mary's Baldock on 4 JANUARY 1863; the third quarter runs JULY TO SEPTEMBER. SO SHE LIVED AT LEAST SIX MONTHS AFTER THE BAPTISM AND HER BOY WAS ABOUT A YEAR OLD WHEN SHE DIED. THE ARCHIVE HAD PUBLISHED, ON /leonard/ AND IN THE LETTER TO A DNA MATCH, THAT 'IF MARTHA GIDDINGS DIED IN 1863 SHE CANNOT HAVE KNOWN HER SON AT ALL'. THAT IS WRONG AND IS CORRECTED. She knew him for the better part of a year, which is a different fact and a sadder one. THE INFERENCE THAT SURVIVES IS THE ONE THAT MATTERED: he is with his grandparents in 1871 and 1881, and a widower of twenty-two took his motherless boy to his wife's people. STILL WORTH ORDERING: the death certificate at 3A/188 gives the AGE AT DEATH, which would confirm the 1841 birth this archive has only from a user tree, and the CAUSE, which would say whether this was childbed fever months later, consumption, or something else entirely. CORRECTION TO THE CORRECTION, SAME HOUR. The commit recording this said the claim was corrected 'in both places', on /tree/ and on /leonard/. IT WAS NEVER ON /leonard/ — that page contains no mention of Giddings at all, and the replace that was supposed to fix it silently matched nothing, which is the same class of bug as the stylesheet link earlier today. ONLY /tree/ CARRIED IT AND ONLY /tree/ NEEDED FIXING. IT IS ALSO IN A LETTER THAT HAS GONE. The reply sent to the second-cousin DNA match this afternoon says: 'If Martha Giddings died in 1863 she cannot have known her son at all, and I notice your George William is with his grandparents in both 1871 and 1881.' The condition is true — she did die in 1863 — so the consequent was asserted, and it is wrong. SHE DIED IN THE THIRD QUARTER AND HE WAS BAPTISED IN THE JANUARY. That goes in the next message to him, not in a fresh email for one sentence. ANSWERED 20 SEPTEMBER, AND THE ANSWER TO THE SECOND HALF IS NO. Three GRO index entries, all Hitchin registration district, in this order: the MARRIAGE of George CASTLE to a GIDDINGS in QUARTER 3 OF 1862; the birth of GEORGE WILLIAM CASTLE in QUARTER 4 OF 1862 - within three months of the wedding; and the death of MARTHA CASTLE in QUARTER 3 OF 1863, volume 3A page 188. SO IT WAS NOT CHILDBIRTH. She died nine to twelve months after her son was born, not at his birth, and the two events are not connected. A second Castle birth in the window, GEORGE MUNT CASTLE of Q4 1863, CANNOT be hers because she was already dead when he was registered - a different Castle family in the same district, written down before it costs anybody a day. THE BOUND, STATED RATHER THAN BURIED: the marriage index prints the groom as George Castle and the bride only as a Giddings - IT DOES NOT GIVE HER FORENAME - and the pre-1866 death index carries NO AGE, so the second Martha Castle death at Hitchin in 1865 cannot be ruled out on the index alone. Closing it properly needs two GRO certificates: the 1862 marriage, which names both fathers, and the 1863 death, which gives her age and cause. WHAT MAKES IT HANG TOGETHER ANYWAY is that George William is with his GRANDPARENTS in 1871 and again in 1881, which is what happens to a boy whose mother died before he was one. In data/castle-giddings-1862.tsv.

IT SETTLES TWO THINGS AT ONCE AND IT IS ONE CERTIFICATE. An England and Wales marriage certificate NAMES BOTH FATHERS with their occupations. So it would give GEORGE CASTLE 1840-1885 independently of a user tree, and it would give CAROLINE'S FATHER — WHICH DECIDES WALLS AGAINST WALLIS, a contradiction this archive has published seven times one way and five the other. The date is October 1882 at Hitchin, and the tree that supplied it notes she 'came to Baldock from Weston just before her first marriage'. NOTE ALSO that their eldest, Sarah Ann, was born 19 August 1882 at Maiden Street, Weston — TWO MONTHS BEFORE THE WEDDING. NOTE 17 SEPTEMBER: THE SPELLING QUESTION IS ALREADY ANSWERED AND NO LONGER NEEDS THIS CERTIFICATE. The GRO birth index gives MOTHER'S MAIDEN NAME WALLIS for six of her children and nothing for Walls. THE CERTIFICATE IS STILL WORTH ORDERING for the other thing it does — it NAMES BOTH FATHERS WITH THEIR OCCUPATIONS, giving George Castle senior independently of a user tree and Caroline's own father for the first time. 17 September: the 1911 census gives YEARS MARRIED as 28, which from 2 April 1911 puts the wedding between April 1882 and April 1883 and so agrees with the Q4 1882 index entry. It is corroboration, not the certificate; the certificate is still what names the two fathers.

A PHRASE IN SOMEBODY ELSE'S TREE. The note against her October 1882 marriage reads 'came to Baldock from Weston just before HER FIRST MARRIAGE'. That wording implies a second, and the arithmetic supports it: GEORGE WILLIAM CASTLE DIED IN 1916 AND SHE LIVED UNTIL 1957 — forty-one years a widow, from the age of fifty-four. NOBODY HAS SEARCHED FOR IT. A remarriage certificate would give her father's name a second time, which is the cheapest possible cross-check on Walls against Wallis, and a second married surname would explain why she has been hard to follow after 1916. FOUND 17 SEPTEMBER, AND THE PHRASE IN SOMEBODY ELSE'S TREE WAS RIGHT. Q3 1920, DISTRICT HITCHIN, VOLUME 3A PAGE 2077: CAROLINE CASTLE MARRIED JAMES JAMES. MarriageFinder names him outright rather than offering four page-mates. She was 57, and George William had been dead four years. THEN THE CONSEQUENCE, WHICH IS BIGGER THAN THE MARRIAGE. FROM 1920 SHE WAS CAROLINE JAMES. Every tree carrying her as 'Caroline Wallis 1862-1957' has been hunting a death registered under a surname she stopped using thirty-seven years before she died. TESTED AND CONFIRMED: England & Wales deaths, CAROLINE JAMES, birth year 1862, died 1957, HITCHIN, HERTFORDSHIRE. AND THE 1939 REGISTER HAS HER, which closes another row in the same stroke: RG101/1598B/017/25, schedule 134, 38 NORTON ROAD, BALDOCK — CAROLINE JAMES, WIDOWED AGAIN, unpaid domestic duties, AND BORN 28 FEBRUARY 1863. AN EXACT DATE OF BIRTH, and it is 1863, NOT the 1862 that this archive and the tree it came from both carry. SHE OUTLIVED TWO HUSBANDS AND DIED AT NINETY-FOUR.

NEW ON 16 SEPTEMBER and nobody has touched him. One of the nine children of George William Castle and Caroline, born at Baldock about 1889, between Ada May 1886 and Agnes Kate 1891. NO DEATH DATE in the tree that supplied him, which usually means he was not followed. WHY HE MATTERS: he is a CASTLE MALE LINE. Descendants of his would carry the Castle surname forward and would share with Derrick through George William and Caroline — and, unlike the Brown and Wear descendants already matched, A MALE-LINE CASTLE COULD ALSO BE Y-TESTED if the question ever needed it. WORKED 17 SEPTEMBER, AND IT RAISES A PROBLEM RATHER THAN CLOSING ONE. A Frederick Castle is in the 1939 Register AT BALDOCK — TNA R39/1598A/017/15, schedule 157 — BORN 10 APRIL 1887, MARRIED, occupation 'VERTICAL DRILLER ON AIRCRAFT WORK', living with ELLA CASTLE born 12 November 1895, AND A THIRD PERSON IN THE HOUSEHOLD WHOSE RECORD IS STILL OFFICIALLY CLOSED. The deaths index carries a FREDERICK JAMES CASTLE, birth year 1887, DIED 1975 AT HITCHIN. Same town, same name, and 1887 in both. THE ARITHMETIC DOES NOT WORK. The tree that supplied the Castle children gives ADA MAY CASTLE AS BORN 10 DECEMBER 1886. A brother born 10 APRIL 1887 IS FOUR MONTHS LATER, WHICH IS IMPOSSIBLE FOR FULL SIBLINGS. THREE READINGS AND NONE IS PICKED. (a) This Frederick is a different Castle — Baldock is small but the surname is not rare there. (b) ADA MAY'S DATE IS WRONG; it comes from the same user tree that put Frederick at 'about 1889' and put George William's birth at Willian when the census says Baldock, so that tree has been wrong twice already today. (c) FREDERICK'S OWN DATE IS WRONG, though a 1939 Register entry is a man stating his own date of birth to a registrar and is about as good as that gets. WHAT SETTLES IT: the birth registration for Frederick James Castle at Hitchin, which would give the quarter and the mother's maiden name — WALLS OR WALLIS — and would either place him in this family or take him out of it. HE STILL MATTERS FOR THE SAME REASON AS BEFORE: if he is theirs, he is a CASTLE MALE LINE and the only Y-testable branch on that side, and the closed record in his 1939 household suggests a descendant. RESOLVED THE SAME HOUR, AND THE TREE WAS RIGHT AFTER ALL. The GRO birth index, searched on MOTHER'S MAIDEN NAME WALLIS at Hitchin, returns FREDRICK JAMES CASTLE REGISTERED IN 1889 — exactly the 'about 1889' the tree gave. SO THE FREDERICK CASTLE BORN 10 APRIL 1887 IN THE 1939 REGISTER AT BALDOCK IS A DIFFERENT MAN, and the four-month impossibility against Ada May dissolves: she is registered 1887 and he is registered 1889, two years apart, which is ordinary. The vertical driller on aircraft work, his wife Ella and the closed record in their household belong to another family. HE IS STILL UNTRACED AND STILL MATTERS: a Castle male line is the only Y-testable branch on that side. What is now known is his registration year and that his death is NOT the 1975 Hitchin one, which belonged to the 1887 man.

UNTOUCHED, AND THE SUBSCRIPTION IS PAID FOR. Findmypast surfaced a hit in THE CITIZEN (LETCHWORTH) against Henry Robert Wear's own record and it was never opened. WHAT A LOCAL PAPER CARRIES THAT NO INDEX DOES: emigration send-offs, works notices, marriage reports naming the whole wedding party, obituaries naming survivors, and court and inquest reports. TARGETS: the Wears at 72 Pix Road 1917-1939 and the 1939 departure; the Castles at Baldock and Norton; J. M. Dent's Temple Press staff notices; and MARTHA GIDDINGS' DEATH IN 1863, which a Hertfordshire paper may report where a certificate is slow and costs money. EXPIRES ABOUT 16 OCTOBER. ATTEMPTED 16 September AND BLOCKED, and the blocker is the interface rather than the subscription. The full-text search at findmypast.com.au/search-newspapers DOES load — it reports 108,257,013 newspaper pages and renders its own form — but IT WILL NOT SUBMIT. The name field accepts a value by every route tried, including a native setter with an input event; the Search button is present and reports as clicked; and the page simply stays where it is with no results and no error. Tried also: the URL forms /search/british-newspapers, /search/newspapers and /newspapers with query strings, ALL OF WHICH REDIRECT to the generic world-records page. SEPARATELY, the indexed FAMILY NOTICES set at search-world-records-in-newspapers-and-publications DOES work and returns NOTHING for WEAR at Letchworth — but that set is not the newspaper archive and says so itself: 'what you are able to search here today is just a small sample of the records to come.' SO THE NEGATIVE IS WORTHLESS AND IS NOT RECORDED AS ONE. WHAT TO TRY NEXT: the Advanced Search link on the same page, which is a different form; britishnewspaperarchive.co.uk directly, which a Findmypast subscription usually covers but which needs its own sign-in; or simply the same search on another day, since the rest of the site has worked all afternoon and this one page has not. OPENED 27 SEPTEMBER 2026, ON DAVID'S OWN SUBSCRIPTION, AND IT PAID ON THE FIRST SEARCH. This row predicted a local paper would carry "marriage reports naming the whole wedding party". IT CARRIED TWO REPORTS OF ONE WEDDING. The CITIZEN (LETCHWORTH) of Friday 6 November 1914 p.4 and the HERTFORDSHIRE EXPRESS of Saturday 7 November 1914 p.4, the second under a BALDOCK heading. Both say Saturday; BOTH RESOLVE TO 31 OCTOBER 1914, where the GRO index gave only "Q4 1914". NEW: the CONGREGATIONAL CHURCH AT BALDOCK, a Nonconformist chapel with the Rev. T. Smith as pastor; MARTHA MAUD CASTLE'S OWN ADDRESS, NORTON STREET, BALDOCK, the first this archive has ever held for her before marriage, and a street in Baldock rather than the village of Norton; that he was "in full uniform", ON FURLOUGH FROM BURY ST EDMUNDS where the 1st Herts Territorials had their headquarters, that the wedding "took place quietly", and that he rejoined and the battalion then had orders for the Front — THE 1/1ST HERTFORDSHIRES LANDED IN FRANCE ON 6 NOVEMBER 1914, the day the Citizen printed its notice, SO HE MARRIED ON ABOUT A WEEK'S LEAVE. AND A NAME VARIANT: the Express calls him HARRY WARE, so a WEAR search does not find him. THE REGIMENT WAS NOT NEWS AND THAT IS THE VALUE — his WO 363 record, number 3010, already gave the 1st Battalion Hertfordshire Regiment, and a service file and a village paper agree without either knowing the other. A SECOND RESULT FROM A DIFFERENT SEARCH: the exact string "72 Pix Road" returns THREE articles in the whole corpus, one a wedding report of 5 October 1935 giving the house to MR AND MRS E. S. CHAPMAN. With the 1921 census (Wears there) and the 1939 Register (18 Ridge Avenue), THE MOVE IS NARROWED FROM 1921-1939 TO 1921-OCTOBER 1935. AND THE THING THIS ROW MOST WANTED IS A NEGATIVE. It hoped a Hertfordshire paper would report "Martha Giddings' death in 1863" and save a certificate. FIRST, THE ROW'S OWN SEARCH STRING WAS WRONG: she died as MARTHA CASTLE, her married name. "martha castle" filtered to county of publication HERTFORDSHIRE returns THREE and none is hers — an 1884 death notice for a Martha Castle aged 39, an 1862 inquest witness called Martha Castle Butler, and an 1872 court list. THE COVERAGE CONTROL HOLDS: the 1862 item is from the Hertfordshire Express, published at Hitchin, so the right paper is digitised for the right period and a notice COULD have been found. But it is weak evidence — a labourer's wife in 1863 usually got no paid death notice — AND IT DOES NOT SEPARATE THE Q3 1863 MARTHA CASTLE FROM THE 1865 ONE. That still needs the certificate. Working in data/wear-castle-wedding-1914-newspapers.tsv.

THE WEARS LEFT IN AUGUST 1939. THE REGISTER WAS TAKEN ON 29 SEPTEMBER 1939, so it catches everyone who did NOT go — and it carries EXACT DATES OF BIRTH, not quarters. TARGETS: Sarah Ann Wallis b.1882, Ada May Castle b.1886, Frederick James b.1889, Florence Margaret Hilda b.1901 and Ernest Albert Cyril b.1903, all of whom died in England; Agnes Kate Castle and Walter George Brown at Blackwater; and CAROLINE HERSELF, then 77 and widowed twenty-three years, whose entry would give her exact date of birth and her household. ALSO WORTH CHECKING WHETHER LEONARD IS IN IT — he is on neither 1939 ship, and an English entry in September 1939 would prove he was still in England a month after his mother sailed. WORKED 17 SEPTEMBER, and it produced two findings and one exclusion. FIRST, CAROLINE. There is no Caroline Castle in Hertfordshire in the 1921 census at all, and there is one CAROLINE JAMES — RG15 piece 7079, schedule 125, 9 BOSCOMBE PLACE, LETCHWORTH, wife of JAMES JAMES, born WESTON, Herts, aged 56 years 4 months. IN THE HOUSE WITH THEM: ERNEST ALBERT CYRIL CASTLE and FLORENCE MARGARITE HILDA CASTLE, ENUMERATED AS STEP-CHILDREN. The remarriage is no longer an index line, it is a household, and the two youngest Castles have their full names for the first time. The age puts her birthday in February, agreeing with the 28th; the year is two out and she gave it herself. AND SHE HAD MOVED TO LETCHWORTH, where her daughter was raising the Wears. SECOND, THE SIBSHIP IS BIGGER THAN ANYBODY THOUGHT. The 1911 census at THE CHURCH YARD, ST MARY'S COTTAGES, BALDOCK — RG14/7572, schedule 173 — asks the fertility question and Caroline answers NINE BORN, EIGHT LIVING, ONE DIED. Six are known from the GRO index and Martha Maud is a seventh. AT LEAST TWO CASTLE CHILDREN HAVE NEVER BEEN IDENTIFIED, and one of the nine was dead before April 1911. JOSEPH WALLIS is in that household as Caroline's father, spelt WALLIS. THIRD, FREDERICK IS NOT SETTLED AND THE 1975 MAN IS NOT HIM: see the row below.

A DIFFERENT WAY AT THE SAME QUESTION, and Findmypast holds the series. Electoral registers list adults at an address year by year, so the year 72 PIX ROAD, LETCHWORTH stops listing Wears is the year the household broke up — which should be 1939 and can now be tested rather than assumed. MORE USEFULLY: LEONARD WAS TWENTY-ONE IN 1938 and would appear in his own right. If he is on the register for 1938 or 1939 and gone by 1940, that dates HIS departure, which BT 27 has so far refused to. WORKED 16 September, AND THE ROW'S OWN PREMISE TURNED OUT TO BE WRONG. It said Leonard 'was twenty-one in 1938 and would appear in his own right', so that the year he vanished would date his departure. HE SAILED IN MARCH 1937, AGED TWENTY — found the same afternoon — AND THE VOTING AGE WAS TWENTY-ONE. He could never have appeared on an English register as an adult, and no amount of searching will put him there. The row cannot do what it was raised to do. WHAT THE REGISTERS DO GIVE IS THE HOUSEHOLD, AND IT IS A SOLID RUN. HENRY ROBERT WEAR IS ON THE LETCHWORTH REGISTER EVERY YEAR FROM 1919 TO 1930 — 1919, 1920, 1921, 1922, 1923, 1924, 1925, 1926, 1928, 1929, 1930, with only 1927 missing from the returned set. Parliamentary County of Hertfordshire, Hitchin Division, polling district LETCHWORTH; British Library SPR.Mic.P.16/BL.H.57. ELEVEN UNBROKEN YEARS is a man rooted in one town, and it brackets the settled period before the emigration. TWO LIMITS, BOTH STATED. The transcripts give the YEAR AND THE POLLING DISTRICT BUT NOT THE STREET — the address is on the original image and has not been read, so this does not yet date the move to 72 Pix Road. AND THE SET ENDS IN 1932, so it CANNOT bracket the 1939 departure from the English end; nothing in it reaches the years that matter for that.

STATED AS A GAP ON /names/ AND STILL OPEN. MOTA FIGUEIREDO is the fullest form anybody ever wrote of Ivone's maiden name — it comes from WLD 1950/1948, the first divorce — and MOTA alone returns 288 records in the archives index, far past the harvest cap used on the other spellings. NOBODY HAS READ THEM. The name is common enough that most will be nothing, but it is the one search where a relative could be hiding in plain sight under a name this archive already knows she used. NEEDS NAAIRS BACK UP — it returned 'the connection to the Search Manager Server is inactive' on the afternoon of 16 September. || BLOCKED 20 SEPTEMBER: NAAIRS has stopped answering altogether — not 503 as on 17 September but no response at all, portal and search service alike, four days running. See row 179. This row cannot move until it comes back, and marking it 'next' overstated what this archive can do. 11 OCTOBER — THE BLOCKER IS GONE. NAAIRS IS BACK AT A NEW ADDRESS UNDER A NEW NAME: https://www.nationalarchives.gov.za/search/, the “National Archival Information Search System”, free, no account, all twenty repositories, with a POST API at /search-api/api/search. The old entry point is still dead, so the service did not come back where this archive was watching. THIS ROW IS NO LONGER WAITING ON AN INSTRUMENT. The instrument's own measurements and the three negatives re-run through it are in data/naairs-is-back-2026-10-11.tsv. 11 OCTOBER — READ IN FULL, AND THE NAME IS A FALSE LEAD FOR A REASON NOBODY HAD GUESSED. All 278 records pulled and read. MOTA IN THE SOUTH AFRICAN ARCHIVES IS OVERWHELMINGLY A SOTHO, TSWANA AND NGUNI SURNAME, NOT A PORTUGUESE ONE. The bulk are NATIVE ESTATE files and magistrates' case files from the Free State and the Witwatersrand — Lehlohonolo, Litaba, Mohasoa, Sophoniah, Mbuti Meshack, Maphehello Agnes, Monamoholo Johane. Depots: TAB 139, SAB 60, VAB 46, NAB 15, KAB 13. EIGHTEEN ARE PORTUGUESE AND NOT ONE IS HERS — an estate (TAB MHG 5905/69, “DA MOTA, JOAO”, the only Portuguese Mota estate in the index), two Cape immigration files, a Portuguese consul-general, two Brazilian diplomats and nine Witwatersrand court files. NONE AT LOURENÇO MARQUES. AND THE TWO-KEYWORD TEST SETTLES IT IN ONE LINE: MOTA and FIGUEIREDO together return EXACTLY ONE RECORD across all twenty repositories — TAB WLD 1950/1948, her own first divorce, held here since September. MOTA+MARQUES zero; FIGUEIREDO+LOURENCO zero. THE ROW IS CLOSED AND THE ANSWER IS NO. In data/naairs-sweep-2026-10-11b.tsv.

FOUND 16 SEPTEMBER BY SPELLING VARIANTS AND ONLY LOGGED, NOT WORKED. Ivone married him on 13 February 1953 and he had sued or been sued at least six times: TPD 467/1950 Rosaline born Bernhardt v him; TPD 2854/1960 him v Rosaline Burns formerly Spence-Wilson formerly Salkow; WLD 1593/1962 him v Gertie formerly Moagus; WLD 3317/1963 him v Cecilia Maud born Shein; WLD 2225/1953 him v Ivonne; and WLD 15934/1979 Mayveen Properties v him. HE MARRIED AT LEAST FOUR TIMES. AND HIS ESTATE PAPERS ARE AT KAB, MOOC 6/9/5171 ref 61416 — CAPE TOWN, where this archive ALREADY HAS AN OPEN REQUEST for the Lerena death notice. IT COULD RIDE ALONG AT NO EXTRA TRAVEL. || BLOCKED 20 SEPTEMBER: NAAIRS has stopped answering altogether — not 503 as on 17 September but no response at all, portal and search service alike, four days running. See row 179. This row cannot move until it comes back, and marking it 'next' overstated what this archive can do. AND A CHECK BEFORE ANYBODY TRAVELS, 20 SEPTEMBER — THIS ROW MAY BE SENDING A RESEARCHER TO THE WRONG PROVINCE. The row proposes that the Cape Town estate KAB MOOC 6/9/5171 ref 61416 'ride along' with the Lerena request. BUT THIS ARCHIVE'S OWN corrections.json SAYS MORRIS ISAAC SALKOW DIED AT JOHANNESBURG: 'Johannesburg Chevrah Kadisha burial register: died 08.11.1991, buried 10.11.1991, WESTPARK BLOCK U GRAVE 2469.' A man who dies and is buried in Johannesburg in 1991 has his estate at the Master there, numbered 1991 or 1992 — NOT at the Cape. AND THERE IS A BETTER CANDIDATE FOR THE CAPE FILE, found in the FamilySearch sweep of 20 September: ISAAC MORRIS SALKOW, born 1897 in LITHUANIA, DIED 10 OCTOBER 1938 IN THE CAPE PROVINCE, with probate 1938 to 1939 in the Cape Probate Records of the Master of the High Court, wife ANNIE DEBORAH born GOLDSMITH. THE FORENAMES ARE REVERSED AND THE DEATHS ARE FIFTY-THREE YEARS APART. MOOC 6/9/5171 fits a 1938 Cape estate far better than a 1991 Johannesburg one. DO NOT ORDER IT AS IVONE'S HUSBAND'S UNTIL THE MOOC REFERENCE HAS BEEN CHECKED AGAINST A DATE. Ivone's husband's estate should be sought at TAB or the Master in Johannesburg for 1991-92. 11 OCTOBER — THE BLOCKER IS GONE. NAAIRS IS BACK AT A NEW ADDRESS UNDER A NEW NAME: https://www.nationalarchives.gov.za/search/, the “National Archival Information Search System”, free, no account, all twenty repositories, with a POST API at /search-api/api/search. The old entry point is still dead, so the service did not come back where this archive was watching. THIS ROW IS NO LONGER WAITING ON AN INSTRUMENT. The instrument's own measurements and the three negatives re-run through it are in data/naairs-is-back-2026-10-11.tsv. 11 OCTOBER — ALL SIX CONFIRMED AND A SEVENTH FOUND. Eighty-nine SALKOW records read. The six this row listed are all there, with two details it did not carry: WLD 1593/1962 gives Gertie as “formerly Moagus, BORN AMBROSE”, and the estate KAB MOOC 6/9/5171 ref 61416 is indexed “SALKOW, ISAAC MORRIS” — the forenames REVERSED from every court file. AND A SEVENTH: TAB WLD 2462/1966, “ISAAC LEWIS BENJAMIN VERSUS ROSALINE EVE (BORN BERNHARDT, FORMERLY (1) BURNS (2) SPENCER-WILSON (3) SALKOW)” — her fourth marriage ending, and the first record to give her a second forename. THE TWO FILES DISAGREE ABOUT THE ORDER OF HER MARRIAGES and this archive picks neither; see litigation.json. In data/naairs-sweep-2026-10-11b.tsv.

OPPOSED APPLICATION. PAULA LUWINSKI VERSUS PHILLIP CRYSTAL, TAB WLD 2451/1972. Found in the variant sweep and not yet explained. THE ARCHIVE KNOWS A PAULINE LUWINSKI — Kurt's widow, born Rappoport 1914, died 1985 — AND PAULA IS NOT PAULINE, or at least is not obviously. The file is from 1972, four years after Kurt died. IF PAULA IS PAULINE UNDER A SHORTER NAME, this is the widow in court over something four years on; IF SHE IS NOT, there is a Luwinski woman in Johannesburg nobody has accounted for. EITHER ANSWER IS WORTH A ROW. It is a TAB file and belongs on any future researcher list. ANSWERED 16 SEPTEMBER FROM THE GAZETTE. Government Gazette 3701 of 10 NOVEMBER 1972, page 126, a table of sequestrations: 'No. of estate / Name and description of estate / Date upon which and division of court by which order made / UPON THE APPLICATION OF'. ENTRY T1301/72: 'PHILLIP CRYSTAL, AN ADULT WHITE MALE, A COMMERCIAL TRAVELLER, MARK HOTEL, O'REILLY ROAD, BEREA, JOHANNESBURG' — order made 17/10/72 in the Witwatersrand Local Division — UPON THE APPLICATION OF PAULA LUWINSKI. So the court file WLD 2451/1972, 'OPPOSED APPLICATION, PAULA LUWINSKI VERSUS PHILLIP CRYSTAL', is a SEQUESTRATION SHE BROUGHT AND WON. She was his creditor; he was the one living in a hotel and he was the one whose estate went. AND THE FIRST READING OF IT WAS WRONG, WHICH IS WORTH RECORDING. The text-extraction of that page returned the fragment 'Paula Luwinski, a commercial traveller, Mark Hotel' — the OCR had jammed two columns of a four-column table into one line, and it read as though PAULA were the commercial traveller living in the hotel. SHE WAS NOT. HE WAS. It took rendering the page as an image at 170 dpi to see the table structure, which is the same technique this archive has needed before on these scans. A COLUMN BOUNDARY IS INVISIBLE TO A TEXT DUMP. WHAT IS NOW KNOWN ABOUT HER: a Luwinski woman in Johannesburg in 1972, four years after Kurt died, solvent enough to be a creditor and determined enough to take an opposed application to judgment. WHAT IS STILL NOT KNOWN is who she was to him — she is not Pauline, whose name appears in every record as Pauline and who was born Rappoport.

HOUSEKEEPING, BUT IT IS LOAD-BEARING. On 16 September the NAAIRS entry point this archive had used throughout stopped resolving, and the service turned out to have MOVED to www.national.archsrch.gov.za/sm300cv/smws/sm300dl. Every reference in litigation.tsv, naairs-figueiredo.tsv and the researcher briefs was taken through the OLD endpoint and none has been re-read through the new one. MOST WILL BE FINE. What matters is that the archive should not cite what it cannot currently reach, AND THE SOURCES PAGE STILL PRINTS THE OLD ADDRESS. Update it, and re-run the citations when the Search Manager Server is answering again. 17 September, a second day and a second probe: the NAAIRS PORTAL answers 200 and THE SEARCH SERVICE ANSWERS 503 AGAIN. Two days, same split. A 503 on two consecutive days is a service that is down rather than a URL that is stale — which is the distinction this archive got wrong on 16 September — but it is still not a finding about any record, and nothing has been published from it. || BLOCKED 20 SEPTEMBER: NAAIRS has stopped answering altogether — not 503 as on 17 September but no response at all, portal and search service alike, four days running. See row 179. This row cannot move until it comes back, and marking it 'next' overstated what this archive can do. || 27 SEPTEMBER 2026: THERE ARE TWO NAAIRS DATABASES, not one that moved. NARSSA's own page says the NEW fulltext database is INCOMPLETE while 8.3 million entries migrate, and that the OLD one gives "a more complete search". Probed today: the old sm300 service 503s, naairs.htm is 404. This row's re-verification job now has a second axis — not just re-reading citations through a working endpoint, but RECORDING WHICH DATABASE ANSWERED. See row 275 and data/naairs-two-databases.tsv. 11 OCTOBER — THE BLOCKER IS GONE. NAAIRS IS BACK AT A NEW ADDRESS UNDER A NEW NAME: https://www.nationalarchives.gov.za/search/, the “National Archival Information Search System”, free, no account, all twenty repositories, with a POST API at /search-api/api/search. The old entry point is still dead, so the service did not come back where this archive was watching. THIS ROW IS NO LONGER WAITING ON AN INSTRUMENT. The instrument's own measurements and the three negatives re-run through it are in data/naairs-is-back-2026-10-11.tsv. 11 OCTOBER — DONE, AND THE WAY IT HAD TO BE DONE IS THE FINDING. A REFERENCE CODE IS NOT SEARCHABLE ON THIS SERVICE AT ALL. “61416” returns ZERO while ZA KAB MOOC_6/9/5171_61416_1 sits in the index, and “534/1948” returns NINE THOUSAND FOUR HUNDRED AND SIXTY-NINE records because the keyword box TOKENISES ON THE SLASH and matches 534 OR 1948 anywhere in a title. The portal searches the TITLE TEXT and the reference code is not in it. SO EVERY CITATION WAS RE-VERIFIED BY SURNAME, which is the only way, and ALL EIGHTEEN CAME BACK WITH THE REFERENCE CODE THIS ARCHIVE ALREADY HELD: TPD 534/1948, WLD 1950/1948, WLD 2545/1950, WLD 497/1951, WLD 2368/1951, WLD 2542/1957, WLD 2451/1972, MHG 7584/69, MHG 4267/57, MOOC 6/9/5171 61416, WLD 2225/1953, TPD 467/1950, TPD 2854/1960, WLD 1593/1962, WLD 3317/1963, WLD 15934/1979, WLD 583/1944, WLD 332/1952. NOT ONE CITATION IN THIS ARCHIVE IS STALE, and the sources page now prints the working address. In data/naairs-sweep-2026-10-11b.tsv.

CHAIN MIGRATION IS A REAL PATTERN AND IT HAS NOT BEEN TESTED HERE. The Dunvegan Castle carried 406 passengers out of London on 10 August 1939 and Martha Maud and four children were five of them. THE QUESTION IS WHETHER ANYONE ELSE ABOARD CAME FROM LETCHWORTH, HITCHIN OR BALDOCK — because people from one small town emigrating in one year usually go together or follow each other, and a neighbour on that manifest would explain WHY Johannesburg of all places, which this archive has never satisfactorily answered. Same for the Durban Castle in April. Findmypast allows browsing a voyage rather than searching a name. WORKED 16 September AS FAR AS THE INDEX ALLOWS, WHICH IS NOT FAR ENOUGH. Findmypast will browse a voyage rather than a name, and the Dunvegan Castle of 10 August 1939 returns FORTY-THREE INDEXED PASSENGERS — not the 406 the manifest states, so the indexed set is a fraction, most likely one class or one sheet. The forty-three are Bliss (two), Catto, Corbett, Coxen, Downwaite, Farber (three), Furneaux, Goldsmith, Griffin, Haward, Hay, Hobling, John, Luslie (two), Macnab and the rest, with the five Wears among them. NOT ONE OF THEM CAN BE PLACED. THE INDEXED FIELDS DO NOT INCLUDE AN ADDRESS — five Wear transcripts from this voyage were read in full earlier today and none carries one — so there is no way from the index to tell whether anybody aboard came from Letchworth, Hitchin or Baldock, which is the entire question. THE ANSWER IS ON THE IMAGE AND ONLY THERE. A BT 27 outbound form carries a column for the passenger's LAST ADDRESS IN THE UNITED KINGDOM, and that column is what would show a neighbour. SO THE ROW IS NOT DONE, IT IS RE-AIMED: view the original manifest images for both voyages — TNA BT27/1542/00/0024 for the Dunvegan Castle and BT27/1540/00/0042 for the Durban Castle — and read the address column beside the Wear entries. That is a reading job, not a search, and it is the only route to why Johannesburg. OPENED 29 SEPTEMBER 2026, AND THE ROW'S OWN PREMISE IS WRONG. This row wanted a Letchworth neighbour on one of the two ships, on the theory that it would answer why South Africa. BT 27 CANNOT ANSWER THAT AND NO AMOUNT OF READING WILL MAKE IT. The 1939 form ends at column (9), 'Country of Intended Future Permanent Residence' — A COUNTRY AND NOTHING FINER. There is no destination address, no consignee, no port-of-call detail per passenger. So the set cannot say where in South Africa anybody was going, and a shared home town would have been suggestive at best. BUT THE PAGES PAID FOR THEMSELVES ON A COLUMN THE ROW WAS NOT LOOKING AT. Column (6), LAST ADDRESS IN THE UNITED KINGDOM, is not in the Findmypast transcript at all and is only on the image. It reads 18 RIDGE AVE, LETCHWORTH for Henry Robert on the Durban Castle of 5 April, and again for Martha Maud and the four youngest — one brace over all five — on the Dunvegan Castle of 10 August. TWO VOYAGES FOUR MONTHS APART, THE SAME HOUSE, and it corrects a published sentence: see row 149. Also read: they entered S AFRICA as intended future permanent residence, so they went as emigrants and not on a passage out and back; and his occupation is MACHINE MINDER in a clear hand, where the index has 'Machine minister'. TWO PAGES WERE READ IN FULL AND TWENTY-FOUR WERE NOT — Durban Castle runs to 11 pages and Dunvegan Castle to 15 — SO THIS IS NOT A VOYAGE-WIDE NEGATIVE AND IS NOT RECORDED AS ONE. On the two pages read: no Hertfordshire address at all on Durban Castle p6, and exactly one on Dunvegan Castle p7, Roundwood Lane HARPENDEN, the household immediately below the Wear brace. Harpenden is fifteen miles from Letchworth and a passenger list is written in booking order; ADJACENCY IS NOT ACQUAINTANCE and the archive is not building anything on a line break. Working in data/wear-1939-bt27-images.tsv. AND THE FORM SURVEY PAID FOR ITSELF THE SAME HOUR. Record Atlas asked whether the no-destination-address finding holds for other BT 27 years. It does not — 1907 has no last-UK-address column at all, 1937 and 1939 are the same P.M. 21 form, and a 1951 piece is a typed US military manifest with none of the Board of Trade form left. THE 1937 HALF OF THAT ANSWER WAS THE USEFUL ONE: it meant Leonard's own Windsor Castle sheet of 12 March 1937 carries column (6), so it was opened. It reads “Wear, Mr L A R · 20 · 18, Ridge Avenue, Letchworth, Herts. · Engineer”. THREE DEPARTURES ACROSS TWENTY-NINE MONTHS, ALL THREE GIVING THE SAME HOUSE, and the street written out in full where both 1939 entries abbreviate it and this archive had read it with a query against the word.

TWO INSTITUTIONAL LETTERS NOBODY HAS WRITTEN. Henry Robert Wear was a PRINTER'S ASSISTANT boarding at 9 Ridge Road in 1911 and a PRINTER'S MACHINE MINDER AT J. M. DENT & SONS at 72 Pix Road in 1921 — the Temple Press, which was the reason Letchworth existed for men like him. DENT'S BUSINESS RECORDS survive in part and staff registers sometimes name date of engagement and previous employer. SEPARATELY, THE LETCHWORTH GARDEN CITY HERITAGE FOUNDATION holds tenancy and estate records for the town's houses — both 9 Ridge Road and 72 Pix Road were Garden City property. A tenancy file would date the family's arrival and departure from the other end, independently of the passenger lists.

THE CENTRAL QUESTION, AND ONE ROUTE TO IT HAS NEVER BEEN TRIED. The archive has written to Mozambique and had no answer, and row 8 records eleven institutional letters still unanswered. WHAT HAS NOT BEEN TRIED IS THE PORTUGUESE SIDE OF THE SAME RECORD. Portuguese colonial civil registration was duplicated centrally, and the CONSERVATÓRIA DOS REGISTOS CENTRAIS IN LISBON holds registrations from the former overseas territories. A Portuguese assento de nascimento names BOTH PARENTS AND ALL FOUR GRANDPARENTS — far more than a South African form ever asked. SHE WAS BORN AT LOURENÇO MARQUES ON 4 MARCH 1924 and died at Porto, so she was a Portuguese national to the end and her record should exist on the metropolitan side.

CLOSED 16 September. Request 18 was drafted and HELD as a separate approach to the researcher about the 1951 ante-nuptial contract. IT NEVER NEEDED TO GO. The contract is in the letter that was actually sent on 16 September — item 3 of three, LEONARD AUBREY ROBERT WEAR and MARJORIE EILEEN DUNLOP born WALKER, certificate 1397 of 1951, flagged to her as THE ONE TO GET IF SHE CAN ONLY GET ONE. Holding it as a separate letter would have been exactly the scattering Ms Xaba complained about, in a different inbox. THE FILE STAYS ON DISK as the record of an approach considered and folded into another.

FOUND 16 September, AND FOUND BY A GUARD RATHER THAN BY LOOKING. Publishing the newly proved Giddins generation made check:matches fail, flagging GIDDINS as a surname belonging to a living match holder. The obvious move was to call it a coincidence and add it to the ambiguous list. IT IS NOT A COINCIDENCE. TWO MATCH HOLDERS CARRY THE SURNAME GIDDINS — 43.36 cM with a twelve-person tree, and 62.89 cM with an eighty-nine-person tree and a public pedigree — AND BOTH SIT IN CLUSTER 12. THE ARITHMETIC FITS. James Giddins of Norton End, born about 1815, is Derrick's THREE-TIMES-GREAT-GRANDFATHER: James to Martha to George William Castle to Martha Maud to Leonard to Derrick. A descendant of James through Martha's brothers GEORGE b.1847 or JOHN b.1849 would be a FOURTH COUSIN, expected around 35 cM on a range that runs to 139. 43 and 63 sit comfortably inside it. WHY IT MATTERS MORE THAN ANOTHER MATCH: every piece of DNA evidence this archive holds for the Castle side runs through GEORGE WILLIAM CASTLE AND CAROLINE — the Brown cousins, the Halsey line, all of it. THESE TWO ARE THE FIRST THAT REACH ABOVE HIM, and they do it on his MOTHER'S side, which no paper record had even identified before this morning. NEXT: read the eighty-nine-person tree and look for George or John Giddins of Norton End; check whether the two Giddins matches also match each other and the Brown cousins, because if they match the Browns too the connection is through Caroline and not the Giddinses at all; and TEST THEIR SEGMENTS against the twelve-region Wear signature, which should show NOTHING — a Giddins cousin shares the Castle side only, so overlap with the Wear regions would mean the reading is wrong. CHECKED WITHIN THE HOUR, AND THE CLAIM ABOVE IS WITHDRAWN. The larger tree — eighty-nine people, a public pedigree — was opened and read. ITS GIDDINS LINE IS WARWICKSHIRE, NOT HERTFORDSHIRE. Its GEORGE GIDDINS, 1846-1890, whose dates sat so neatly against our George b.1847, WAS BORN ON 17 JULY 1846 AT WITHYBROOK, WARWICKSHIRE and died in the Wandsworth district; HIS FATHER IS RICHARD GIDDINS, NOT JAMES. The surrounding surnames in that tree — Ratcliffe, Conway, Woodhams, Wall, McMellon, Prescott — are nowhere near Baldock. SO THE SURNAME IS NOT THE CONNECTION. The match is real — 62.89 cM is 62.89 cM — but a match holder's surname is only one of eight great-grandparental lines and there is no reason it should be the relevant one. WHAT I DID WRONG, AND IT IS WORTH NAMING: I saw a surname I had discovered that morning, found dates that fitted, and wrote three paragraphs about 'the first DNA to touch the Castle line above George William' BEFORE OPENING THE TREE. The arithmetic I did was sound and the premise was never checked. It took four minutes to check and it was wrong. WHAT IS ACTUALLY LEFT: two real matches at 43.36 and 62.89 cM in cluster 12, connected to Derrick for some reason that is not the Giddins surname; the smaller tree (twelve people) is too thin to read and has not been. THE HERTFORDSHIRE GIDDINSES REMAIN UNTOUCHED BY ANY DNA.

DONE 16 September on David's instruction, with the painting he had sent days earlier and the Defranceschi site as the reference. THE PLUMBING WAS ALREADY THERE AND NOBODY HAD NOTICED: index.astro has looked for hero.jpg in site/public since it was written, and its alt text ALREADY DESCRIBED THIS EXACT PAINTING — a railway running away from the viewer, winter in the foreground, dry highveld beyond, a river town at the vanishing point. The image had simply never been dropped in. THE IMAGE: 3.2 MB PNG down to 500 KB JPEG at 1800px, which is the difference between a header and a penalty. THE HERO now follows their pattern rather than sitting under it — a gradient that darkens only the lower third so the sky stays open, a gold letterspaced kicker, the title set OVER the plate, and the page's own h1 suppressed so it is not printed twice. Caption below the band, unchanged: 'An illustration, not a document. Nothing in this archive rests on it.' THE HEADLINE WAS REWRITTEN AFTER LOOKING AT IT. The first version ran to four lines and covered the whole painting; the Defranceschi one is a short concrete claim, so this is now OUT OF EAST PRUSSIA AND MOZAMBIQUE, INTO ONE JOHANNESBURG MARRIAGE, which is three lines and leaves the vanishing point visible. THE MASTHEAD: sticky bar, serif wordmark with ARCHIVE in italic accent, and TWENTY-THREE FLAT LINKS REPLACED BY FIVE GROUPED DROPDOWNS — Start, Our line, Places, People & names, Evidence — with a search pill, and a Menu drawer under 62rem. AND THE PHONE WAS CHECKED AND WAS WRONG: at 375px the caption block had covered the entire painting and there was no picture left. Image height raised to 26rem, headline down to 1.5rem, crop shifted left to 42% so the winter side stays in frame.

DONE 16 September on David's instruction to bring this site closer to the Mazza and Defranceschi archives. WHAT THE INVESTIGATION FOUND: those two share a HOUSE STYLESHEET — identical variable names, identical class names, differing only in palette — and a COMPONENT LIBRARY in @daviddef/archive-kit that this archive already depends on and was using for exactly one thing, the work list. The kit ships SiblingArchives, Evidence, AncChart, EraTimeline, SovereigntyTimeline, MarriageChart, Distribution, Numbers, Spine, Search, Atlas, Bloodline and more. WHAT WAS ADOPTED: the house stylesheet, repalletted to this archive's own rust and slate rather than shipping a second set of colours; the three house typefaces, EB Garamond, Libre Franklin and IBM Plex Mono; and ALIAS TOKENS so shared component CSS resolves against the existing palette instead of duplicating it. The grouped dropdown nav and the hero were already done earlier the same day. AND WHAT WAS TAKEN BACK OUT, WHICH MATTERS MORE. I added this archive to the ring of sibling archives and wired the component into the footer. DAVID: 'i dont want the other archives on the luwinski page. not related.' HE IS RIGHT, AND IT IS A CORRECTNESS POINT RATHER THAN A PREFERENCE. The ring's own sentence reads 'each following a different line of THE SAME FAMILY'. The seven are David's own lines — Istria, Kvarner, Carnia, Sicily, Calabria, the Cape, Ireland. THIS FAMILY IS HIS WIFE'S FATHER'S, out of East Prussia and Portuguese East Africa, and it meets the others nowhere. Publishing it would have made all eight sites assert a relationship that does not exist, in a sentence none of them would have re-read. Removed from the layout and from the kit's own archives.json, so the other seven do not start claiming him either. ONE CHANGE KEPT: the component now reads its count from the file rather than saying 'one of seven' in words, which was wrong the moment an eighth was added and would have been wrong again.

DONE 16 September. The tree page had four hand-rolled marks — proved, inferred, unknown, excluded — invented that morning. The sibling archives share a FIVE-WORD LADDER in @daviddef/archive-kit: DOCUMENTED, a record was read and it says this; PROBABLE, the evidence points one way and nothing points the other; INFERRED, reasoned from what is known and not witnessed in a record; FAMILY, told or written on somebody's chart and not sourced; DISPUTED, two sources disagree and neither has been ruled out. The page now uses the shared component, so a level means the same thing wherever a reader meets it. TWO OF THE OLD FOUR WERE NEVER EVIDENCE GRADES AND ARE NOT FORCED ONTO THE LADDER. 'Unknown' is the ABSENCE of a claim; 'excluded' is a claim TESTED AND FAILED. Both keep their own chips, drawn deliberately unlike the ladder — no fill, dashed edge — so they cannot be mistaken for a grade, and the legend says in capitals that they are not levels. AND THE EXERCISE COST ONE CLAIM, WHICH IS THE REAL VALUE OF IT. GEORGE WILLIAM CASTLE AND CAROLINE WALLIS WERE MARKED PROVED AND ARE NOW MARKED PROBABLE. The marriage, the nine children and the dates came from another researcher's tree, not from records this archive has read. What IS documented, because the 1871 census was pulled and read here, is that a George Castle, 30, agricultural labourer, was at Norton End as SON-IN-LAW to James Giddins with a son George, 8. What is NOT is the birthplace at Willian, the October 1882 marriage and the nine children. THE CENSUS ALSO DISAGREES WITH THAT TREE TWICE — it gives the boy's birthplace as BALDOCK, not Willian, and his birth year as 1863, not 1862. Under the ladder that is exactly what PROBABLE means. The marriage certificate, already on the list, would settle it.

DONE 16 September on the Findmypast subscription. The 1939 Register was taken on 29 SEPTEMBER 1939, seven weeks after the Dunvegan Castle left London — so it catches exactly the people who did not go. SEARCHED WEAR AT LETCHWORTH: ONE RESULT, AND IT IS HER. TNA R39/1620/1620C/004/42, 18 RIDGE AVENUE, LETCHWORTH, schedule 279: LOUISA CORBY b. 14 Jan 1870, single, unpaid domestic duties; LOTTIE CORBY b. 3 Mar 1881, single, hand sewer; and JEANNETT (E M) WEAR, B. 28 OCTOBER 1919, SINGLE, CINEMA CASHIER. AN EXACT DATE OF BIRTH WHERE THIS ARCHIVE HAD ONLY 'JANETTE 1919'. AND SHE IS NOT WITH FAMILY. She is nineteen, lodging with two spinster sisters, and working in a cinema. Her mother and the four youngest had sailed for Durban on 10 August; her father had gone in April. SHE STAYED, WITH A JOB AND A ROOM. Note also RIDGE AVENUE against the 9 RIDGE ROAD where her father boarded in 1911 — the same corner of the same small town, twenty-eight years apart. LEONARD IS NOT IN THE REGISTER AT LETCHWORTH OR HITCHIN, which is consistent with his having already gone, BUT IT IS NOT PROOF: the register's own notice says officially closed records are withheld for people not proven dead, and LEONARD'S DEATH YEAR IS PRECISELY WHAT THIS ARCHIVE CANNOT SETTLE — 1973 or 1979. If the National Archives does not know he died, his entry would be closed and would look exactly like absence. A search on WEAR born 1916-1918 returns 61 people and none is a Leonard. Saved to data/wear-1939-register.tsv. CORRECTED 29 SEPTEMBER 2026: THE WORD 'LODGINGS' WAS THIS ARCHIVE'S INVENTION. The register gives an address, not a tenure. Because the family's known house was 72 Pix Road, this row assumed 18 Ridge Avenue must be a room she had taken, and the phrase 'in lodgings' went onto Leonard's page and into this row's own title. THE PASSENGER LISTS SETTLE IT: 18 RIDGE AVENUE IS THE FAMILY'S LAST ADDRESS IN THE UNITED KINGDOM ON BOTH 1939 VOYAGES. It was their house. She had not moved out and taken a room half a mile from her grandmother; she was alone in the house her father left in April and her mother and the four youngest left in August. The distance from 63 Pix Road is unchanged and so is everything about the marriage in December; what changes is whose roof she was under, and the archive had filled that in without a record. Registered in data/superseded.json as janette-ridge-avenue-lodgings. AND STRENGTHENED THE SAME DAY: Leonard's 1937 list gives 18 Ridge Avenue too, so it was the family address from at least March 1937 through August 1939 — not an address they happened to be at when they left.

SURFACED 16 September against the BT 27 transcript of Leonard's 1937 crossing and not yet read. FOUR RECORDS: L A R WEAR, GERMANY 1939; and L A R WEAR, GREAT BRITAIN, 1943, 1944 and 1945. TREAT THEM AS A PROBLEM BEFORE TREATING THEM AS A LEAD. A man who sailed from Southampton to the Cape in March 1937 and married at Rosebank, Johannesburg, in September 1940 is not obviously in GERMANY IN 1939, and the 1943-45 Great Britain records sit oddly against a life this archive believes was lived in Johannesburg. THREE READINGS, AND THEY ARE NOT EQUALLY LIKELY. (a) THEY ARE A DIFFERENT L A R WEAR — the initials are not rare and Findmypast's suggestions match on name alone, which is exactly how the Giddins error happened this afternoon. (b) THEY ARE HIM, AND HE CAME BACK — possible, and it would mean the 1937 emigration was not permanent. (c) THE 'GERMANY 1939' RECORD IS A PRISONER-OF-WAR OR CASUALTY INDEX whose date is the campaign rather than the man. WHAT TO DO: open each one and read the service number, the unit and any date of birth BEFORE deciding it is him. If any carries a birth date matching Q1 1917 at Hitchin, that settles it. THIS ALSO BEARS ON ROW 25, the SANDF service record, which assumes he served South African. READ 16 SEPTEMBER, AND THEY ARE HIM. The row said to treat them as a problem before a lead and to read the service number before deciding anything. THE SERVICE NUMBER IS WHAT DECIDED IT. All five records carry 141272; the rank is CORPORAL; and the force is not a British regiment but the SOUTH AFRICA UNION DEFENCE FORCES. TNA WO 392/15 and WO 392/21, the War Office Directorate of Prisoners of War lists for 1943-1945. HE WAS A PRISONER OF WAR, HELD IN GERMANY, listed in 1943, 1944 and 1945. THE 'GERMANY 1939' TAG THAT LOOKED SO WRONG IS THE CAMP, NOT A POSTING, and the 1939 is the collection's war-span label rather than a date. THE SOUTH AFRICAN FORCE IS WHAT TURNS INITIALS INTO A PERSON. Leonard sailed for the Cape in March 1937, married at Rosebank in September 1940, and was twenty-five when Tobruk fell in June 1942 and something over ten thousand South Africans were taken prisoner in a day. AND IT EXPLAINS THE SILENCE THIS ARCHIVE HAS BEEN RECORDING FOR WEEKS. Row 41: Leonard cannot be placed between 1941 and 1950 — no address, one recorded fact, a daughter born in 1941. A MAN IN A PRISON CAMP DOES NOT APPEAR IN A DIRECTORY, ON A VOTERS' ROLL OR IN A NEWSPAPER. The hole is not an absence of records; it is an absence of liberty. HELD AS PROBABLE, NOT DOCUMENTED: the lists give INITIALS ONLY AND NO DATE OF BIRTH, so the case rests on initials, force, a fitting age, and one number repeating across five records in two volumes. WHAT SETTLES IT IS THE SANDF SERVICE RECORD AT PRETORIA — row 25, wanted since the archive began and never askable precisely. IT CAN NOW BE ASKED FOR BY NUMBER.

DRAFTED 16 September, NOT SENT. A FOLLOW-UP INSIDE THE EXISTING THREAD 1a0a3b70c6389154, not a new letter — the Centre already has the 15 September enquiry and a second approach would be the scattering Ms Xaba complained of, in a different inbox. WHAT MAKES IT WORTH SENDING IS THAT IT REMOVES WORK RATHER THAN ADDING IT. The original letter asked, at question 3, whether the medal rolls might yield a force number, and said in as many words that WITH A NUMBER THE REST BECOMES ORDINARY. The number is now in hand: 141272, Corporal, South Africa Union Defence Forces, prisoner of war in Germany, from TNA WO 392/15 and WO 392/21. SO QUESTIONS 3, 4 AND 5 OF THE ORIGINAL ARE WITHDRAWN — no medal-roll search, no nominal index needed. THE CAVEAT IS STATED RATHER THAN BURIED: the POW lists carry initials only and no date of birth, so the identification rests on initials, the South African force, a fitting age and one number repeating across five records. The letter says plainly that 141272 may belong to somebody else and that being told so would be a useful answer. FOUR ASKS: the attestation or personnel record for 141272; whatever it gives for PARENTS OR NEXT OF KIN, which is still the heart of it because South African marriage registers of the period have no parents' column and an attestation form does; the PRISONER-OF-WAR AND REPATRIATION PAPERS, which are sometimes filed separately and are worth asking for now rather than in a second exchange; and THE ADDRESSES GIVEN ON ENLISTMENT AND DISCHARGE, because a discharge address is the first thing that would place him after the war and 1948 is the year that matters. SENT 16 SEPTEMBER 2026, message 1a0aa27caa5e8b60, inside the existing thread 1a0a3b70c6389154. The chase clock on that thread is reset to a fortnight. IF THEY ANSWER AT ALL, THE ATTESTATION FORM IS THE MOST IMPORTANT SINGLE DOCUMENT THIS ARCHIVE HAS EVER ASKED ANYBODY FOR: the entire Wear chain rests on Leonard being a son of Henry Robert Wear and Martha Maud Castle, and the only source that says so is a seven-person tree built by his own daughter. An attestation form naming his parents would turn that into a government record.

FOUND MISSING 16 September while going deep on Kurt at David's instruction, AND IT IS THE MOST SERIOUS OMISSION ON THE PRETORIA LIST. Twenty-six references went to the researcher this afternoon — seven Booyzen, seven Wear and Luwinski court files, five Lerena, and the six already agreed — AND KURT'S OWN ESTATE FILE IS NOT AMONG THEM. It has been in this archive's own records for days: ZA TAB MHG_#_7584/69, LUWINSKI, KURT, the estate opened after he died at Johannesburg on 15 December 1968. WHY IT MAY BE THE MOST IMPORTANT MHG FILE IN THE WHOLE ARCHIVE. A South African death notice, form J294, asks THREE things this archive wants and cannot otherwise get. (1) THE DECEASED'S PARENTS. Kurt's father was JACOB LEIBHOLZ, deported to Auschwitz in February 1943, and his mother AUGUSTE LUWINSKI — who did not marry until January 1924, TWELVE YEARS AFTER KURT WAS BORN, which is why he stayed a Luwinski while his younger brother was registered a Leibholz. THE ARCHIVE ASSERTS THAT PARENTAGE FROM GERMAN SOURCES AND HAS NO SOUTH AFRICAN DOCUMENT FOR IT. The J294 would be one, signed by whoever buried him. (2) HIS PLACE OF BIRTH, which would test the East Prussian story against a form filled in by his family rather than by him. (3) AND THE ONE THAT MATTERS MOST: THE LIST OF SURVIVING CHILDREN. Derrick was legally Kurt's son and was about twenty when Kurt died. WHETHER THE DEATH NOTICE NAMES HIM — AND HOW — IS A FACT ABOUT WHAT THE FAMILY BELIEVED AND RECORDED IN 1968, twenty years after the marriage that produced him and fifty-eight before the DNA said otherwise. EITHER ANSWER IS WORTH HAVING. HOW IT SHOULD GO, AND NOT BY A THIRD EMAIL TODAY. The researcher has just been sent a long list and told to cut from the bottom; a separate message adding a twenty-seventh item would be exactly the scattering Ms Xaba complained of, in another inbox. SHE HAS NOT YET SENT THE INVOICE, so there will be a natural reply. IT RIDES ON THAT ONE, and it should be marked as belonging near the TOP, not the bottom.

TESTED AND ELIMINATED 16 September, and the elimination is the useful part. Kurt arrived in South Africa in 1933, a German Jew of twenty-one leaving the year Hitler took power, and THIS ARCHIVE HAS NEVER ESTABLISHED HOW HE TRAVELLED. The obvious test was BT 27, the British outbound passenger lists, because a great many refugees went out through Southampton. SEARCHED ON LUWINSK* ACROSS THE WHOLE SERIES 1890-1960: SIX RESULTS, ALL ONE POLISH FAMILY OF LUWINSKAS SAILING LIVERPOOL TO PORTLAND, MAINE, IN 1908. He is not in it. SO HE DID NOT LEAVE FROM A BRITISH PORT, and the record of his crossing is not at Kew. THAT POINTS SOMEWHERE SPECIFIC: the HAMBURGER PASSAGIERLISTEN, the Hamburg passenger lists, which run to 1934 and are held at the Staatsarchiv Hamburg — or the Bremen lists, most of which were destroyed. A Hamburg entry would give his age, his last residence, his occupation and his destination port IN HIS OWN EMIGRATION YEAR, which is the one document that would show what he left with and what he called himself. NOT REACHABLE FROM THE CURRENT SUBSCRIPTIONS: the Hamburg lists are indexed on Ancestry, which is signed in but not subscribed, and at the Staatsarchiv direct. A single pay-per-view or a trial month would cover it, which is the same narrow case already recorded for the WWI pension records. BOUNDED 17 SEPTEMBER RATHER THAN ANSWERED. The HAMBURGER PASSAGIERLISTEN run from 1850 TO 1934, so 1933 IS INSIDE THEM — the dates are right and the idea is sound. THE PROBLEM IS ACCESS. They are in practice an ANCESTRY-HELD dataset, and this archive is signed in to Ancestry but NOT SUBSCRIBED. A FAMILYSEARCH ROUTE WAS TRIED AND A FALSE NEGATIVE WAS CAUGHT BEFORE IT WAS BELIEVED. Searching what was taken to be the Hamburg collection for LUWINSKI returned zero — and then the same collection was searched for SCHMIDT, THE COMMONEST GERMAN SURNAME, AND ALSO RETURNED ZERO. THE COLLECTION IDENTIFIER WAS WRONG AND THE FIRST ZERO MEANT NOTHING. That is the third time in a week a control has stopped this archive publishing an empty search as a finding, and it is the reason the control is now run first. WHAT IS LEFT: an Ancestry subscription, which is a purchase and not mine to make; or the Staatsarchiv Hamburg direct. AND A NARROWING WORTH KEEPING: the archive has already excluded Britain through BT 27, and BREMEN'S PASSENGER LISTS WERE ALMOST ALL DESTROYED, so Hamburg is not merely the best remaining door — it is close to the only one. || BLOCKED 20 SEPTEMBER, and the bound is the same one row 119 describes. The Hamburger Passagierlisten run 1850-1934, so 1933 is inside them and the reasoning holds. In practice they are an Ancestry-held dataset and this archive is signed in there but NOT SUBSCRIBED. A FamilySearch route was tried and produced a false negative that a SCHMIDT control caught before it was believed. Nothing further can be done from here without the subscription or a direct order to the Staatsarchiv Hamburg.

FOUND 16 September going deep on Kurt. His last wife PAULINE, born RAPPOPORT 1914, died 1985, is carried in the register as 'Pauline Rappoport / Miller (Luwinski)' and nobody had asked what the Miller was. THERE ARE TWO EARLIER MARRIAGES AND BOTH ENDED IN COURT. TAB WLD 583/1944 — ILLIQUID CASE, DIVORCE, PAULINE MILLER (BORN RAPPOPORT) VERSUS JOSEPH. TAB WLD 332/1952 — ILLIQUID CASE, DIVORCE, PAULINE BERMAN (FORMERLY MILLER, BORN RAPPOPORT) VERSUS ISRAEL. So she married JOSEPH MILLER, divorced him in 1944, married an ISRAEL BERMAN, divorced him in 1952, and married Kurt after that. BOTH REFERENCES ARE NEW TO THIS ARCHIVE and are added to litigation.tsv. AND IT COMPLETES A PATTERN NOBODY HAD NAMED. Kurt married at least three times and EVERY WIFE HAD BEEN MARRIED BEFORE: Ivone was a de Witt when he married her in June 1948 and had been married four months; Frances Hostowsky appears in the records as Hostowsky Berner Sefton Luwinski, which is three surnames before his; and Pauline arrived with two divorces behind her. A man who married three much-divorced women in a decade in one city is a fact about the man, not just about them. PRIORITY LOW AS FILES — they are Pauline's marriages, not Kurt's, and neither would name his parents — but they belong on any future Pretoria list and they are cheap to add to one. || BLOCKED 20 SEPTEMBER: NAAIRS has stopped answering altogether — not 503 as on 17 September but no response at all, portal and search service alike, four days running. See row 179. This row cannot move until it comes back, and marking it 'next' overstated what this archive can do. 11 OCTOBER — THE BLOCKER IS GONE. NAAIRS IS BACK AT A NEW ADDRESS UNDER A NEW NAME: https://www.nationalarchives.gov.za/search/, the “National Archival Information Search System”, free, no account, all twenty repositories, with a POST API at /search-api/api/search. The old entry point is still dead, so the service did not come back where this archive was watching. THIS ROW IS NO LONGER WAITING ON AN INSTRUMENT. The instrument's own measurements and the three negatives re-run through it are in data/naairs-is-back-2026-10-11.tsv. 11 OCTOBER — BOTH CONFIRMED AND THERE IS NO THIRD. Seventy-five RAPPOPORT records across all twenty repositories, and exactly two are hers: TAB WLD 583/1944 “PAULINE MILLER (BORN RAPPOPORT) VERSUS JOSEPH” and TAB WLD 332/1952 “PAULINE BERMAN (BORN RAPPOPORT) (FORMERLY MILLER) VERSUS ISRAEL”. Joseph Miller divorced 1944, Israel Berman divorced 1952, Kurt after that. NO FOURTH HUSBAND AND NO FIFTH on this index, and the pattern this row named — every wife of Kurt's had been married before — survives the check. In data/naairs-sweep-2026-10-11b.tsv.

NOTICED 16 September while looking for the executor of MHG 7584/69. South African deceased estates are ADVERTISED IN THE GOVERNMENT GAZETTE — this archive has used exactly those notices to find the Wear estates at Cape Town and the two Figueiredo estates — and the notice names the executor and gives an address for claims. KURT DIED ON 15 DECEMBER 1968 AND HIS ESTATE IS MHG 7584/69, SO A NOTICE SHOULD APPEAR IN 1969 OR 1970. SEARCHED THE GAZETTE FOR LUWINSKI AND LUWINSKY ACROSS 1968 TO 1972: ONE HIT ONLY, and it is the 1972 sequestration brought by Paula Luwinski. NOTHING FOR KURT AT ALL. THREE READINGS AND THE ARCHIVE SHOULD NOT PICK ONE YET. (a) The OCR lost it — entirely possible on these scans, and this archive has already seen 'Figueiredo' rendered as 'Epo .Figueiredo,CarlosDe'. (b) The estate was small enough to be wound up without the advertisement that larger estates require. (c) The notice ran under a spelling nobody has tried. WHAT IT MEANS PRACTICALLY: the cheap route to Kurt's executor is closed, and MHG 7584/69 itself — already raised as the most serious omission from the Pretoria list — becomes the only way to the death notice and the children it names. CONTROLLED 20 SEPTEMBER, AND ONE OF THE THREE READINGS IS NOW DEAD. The row offered three explanations and said the archive should not pick one. THE FIRST CAN NOW BE ELIMINATED. Reading (a) was that the gazette corpus might not cover 1969 and 1970 properly. IT DOES: searching those years for the phrase every estate notice carries returns 221 issues in 1969 and 163 in 1970 for 'Estates of deceased persons'; 100 and 93 for the Afrikaans 'boedel'; 111 and 101 for 'Master of the Supreme Court'. THE GAZETTES ARE THERE, THEY ARE INDEXED, AND THEY ARE FULL OF ESTATE NOTICES IN EXACTLY THE YEARS HIS SHOULD HAVE APPEARED. Meanwhile LUWINSKI and LUWINSKY return the same EIGHT documents across the entire corpus — three of them the 1942 dissolution of the Oppenheimer partnership, two the 1941 transfer that created it, and one the 1972 Paula Luwinski sequestration — AND NOTHING AT ALL IN 1968, 1969, 1970 OR 1971. The nearest thing to a near-miss was chased and is not him: 'Lewinski' returns one hit in 1969 and one in 1970 and both are VON LEWINSKY in what appears to be a tariff list. READING (b) CANNOT BE ELIMINATED AND THE REASON IS MEASURABLE: the search does no prefix matching — 'Luwinsk' returns ZERO where 'Luwinski' returns eight — so a name the scanner mangled is simply invisible, and this archive has already watched this corpus render Figueiredo as 'Epo .Figueiredo,CarlosDe'. SO IT IS DOWN TO TWO: the OCR lost him, or no notice was ever published. NEITHER CAN BE SETTLED FROM OUTSIDE. What settles it is the estate file itself, TAB MHG 7584/69, which is row 152 and is not on the researcher's list.

RAISED 16 September. The reply sent to him this afternoon contains one sentence that later research contradicts: 'If Martha Giddings died in 1863 she cannot have known her son at all, and I notice your George William is with his grandparents in both 1871 and 1881.' THE DEATH INDEX WAS THEN FOUND — third quarter 1863, Hitchin, volume 3A page 188 — AND HIS BAPTISM IS 4 JANUARY 1863. She lived at least six months past it; the boy was about a year old. SHE KNEW HIM. WHAT TO SAY AND WHAT NOT TO. It is one sentence in a long letter, it was offered as an observation about HIS tree rather than a claim about mine, and he has given no sign of caring either way. IT DOES NOT WARRANT AN EMAIL OF ITS OWN — a correspondent who sends a second message to amend a subordinate clause is a correspondent people stop answering. IT GOES IN THE NEXT REPLY, briefly, with the volume and page so he can check it, AND WITH THE PART THAT IS BETTER THAN AN APOLOGY: the finding itself is new to both trees, and it is his ancestor as much as ours. DRAFTED 17 SEPTEMBER, NOT SENT: requests/27-SEND-THIS-to-paul-correction.txt. It leads with the correction rather than burying it. THE ERROR: 'If Martha Giddings died in 1863 she cannot have known her son at all' — she died in the THIRD QUARTER of 1863 and he was baptised on 4 JANUARY of it, so she had him for something under a year. Both numbers were already on file and the quarters were never compared. IT ALSO CHANGES THE READING OF THE 1871 CENSUS, which the letter says: the widower is IN the house at Norton End, enumerated son-in-law with his boy as grandson, so he did not give the child away — he went to his dead wife's parents and was still there eight years later, and those two relationship words are the census's own and are now the proof of the line. A second caution goes with it: the same census gives the boy's birthplace as BALDOCK not Willian and his birth year as 1863 not 1862, and the letter says so rather than quietly amending his tree. THEN THREE THINGS THAT ARE HIS, not ours: CAROLINE REMARRIED — Q3 1920 Hitchin, Caroline Castle to JAMES JAMES — and was CAROLINE JAMES when she died in 1957, which is the whole reason his great-grandmother cannot be followed past 1916; her exact date of birth is 28 FEBRUARY 1863 from the 1939 Register at 38 Norton Road, not the 1862 every tree carries; and the surname is WALLIS not Walls, six GRO birth entries to none. It closes with Pix Road, flagged as not his line but worth having. Offers images of everything. THE ERROR IS NOW NAMED EXACTLY, 20 September. The drafted letter says "If Martha Giddings died in 1863 she cannot have known her son at all." THAT SENTENCE IS WRONG. George William Castle was registered in Q4 1862 and Martha Castle died in Q3 1863, so SHE HAD HIM FOR ABOUT NINE MONTHS. The letter is still unsent, so this is a correction to a draft rather than a correction in public - but it is the kind that has to go in before the envelope, not after. The rest of the Martha Giddings paragraph stands, including the reading that a granddaughter born thirty-two years later was christened Martha Maud for her.

FOUND 17 September, and it closes three separate questions at once. EMMA HAWKINS, WIDOWED AT THIRTY-TWO IN 1891, HAD NEVER BEEN FOLLOWED PAST THAT CENSUS. Three direct tests failed: no Emma Wear in Hertfordshire in the 1901 census, no remarriage indexed under Emma Wear, no death. SHE WAS SEARCHED FOR UNDER THE WRONG SURNAME. Coming at it through the children found her at once. 22 MOUNT PLEASANT, ST ALBANS — 1901 census, RG13/1312: JOHN HULL, head, 33, RAILWAY LABOURER; EMMA HULL, wife, 41, BORN ALDENHAM — which is what identifies her, since Aldenham is a small parish and it is her birthplace in both the 1861 and 1871 returns. With ALBERT J HULL aged 6 and EDITH M HULL aged 4, both born St Albans; and LOUISA S A WEAR 18, ROSETTA E WEAR 15, AND HENRY R WEAR 11, ALL ENTERED AS STEP-CHILDREN OF JOHN HULL. WHAT IT SETTLES. (1) Emma remarried about 1893-94 and had two more children. Henry Robert was raised from about four by a stepfather who was a railway labourer. (2) ROSETTA WAS ALIVE AT FIFTEEN and is an APPRENTICE STRAW HAT MAKER — the Hertfordshire county trade, the same industry in which her step-grandmother Elizabeth Giddins was doing 'platting' at Norton End in 1871. She had simply been searched for under Wear in the wrong decades. (3) LOUISA IS CONFIRMED AS LOUISA S A WEAR, which is exactly the form that married a SHENTON at Hitchin in 1918. AND IT OPENS SOMETHING NEW AND DNA-RELEVANT: ALBERT J HULL b.1895 AND EDITH M HULL b.1897 ARE HENRY ROBERT'S HALF-SIBLINGS. Their descendants would share with Derrick THROUGH EMMA HAWKINS ONLY — not through Robert Wear and not through the Castles — which isolates the Hawkins line by itself. No match on the current list carries Hull.

FOUND 17 September while looking for Louisa's Shenton household, and it is the most striking thing the Findmypast month has produced. TNA R39/1620/1620G/011/28, SCHEDULE 305, 63 PIX ROAD, LETCHWORTH, taken 29 September 1939: EMMA HULL, BORN 7 FEBRUARY 1859, WIDOWED, 'INCAPACITATED - BLIND'; LOUISA SHENTON, born 29 November 1882, married; WILLIAM SHENTON, born December 1876, 'INCAPACITATED'. EMMA HULL IS EMMA HAWKINS. Born at Aldenham, married Robert Wear at Southwark in 1881, widowed in 1890, remarried John Hull about 1893, AND WIDOWED A SECOND TIME BY 1939. At eighty she is blind and living with the daughter who was eight when her father died. HER EXACT DATE OF BIRTH IS NOW KNOWN and it comes from the best kind of source — a woman of eighty giving it to a registrar in person. SO IS LOUISA'S: 29 NOVEMBER 1882. AND THE ADDRESS IS THE FINDING. HENRY ROBERT WEAR AND MARTHA MAUD RAISED SIX CHILDREN AT 72 PIX ROAD. HIS MOTHER AND HIS SISTER WERE AT 63 PIX ROAD. THE SAME STREET, a few doors along, in a garden city of eight thousand that had not existed forty years earlier. IT GIVES THE EMIGRATION A SHAPE IT DID NOT HAVE. Leonard went in March 1937, his father in April 1939, his mother in August with the four youngest — and the register, taken seven weeks after that ship sailed, finds still in Pix Road an eighty-year-old blind woman and the daughter keeping her, with Janette in lodgings half a mile off on Ridge Avenue. THE RECORD SAYS WHO WAS WHERE AND NOTHING ABOUT WHAT ANYBODY FELT, and the page says so. BUT IT IS NO LONGER TRUE THAT THE FAMILY SIMPLY LEFT. Still open: no children are in the Shenton household, so whether Louisa had any — the independent test of the Wear half — is unanswered; she was 57 and any would have gone.

OPENED 17 SEPTEMBER by the fertility question on the 1911 census. RG14/7572, schedule 173, St Mary's Cottages, Baldock: CHILDREN BORN 9, STILL LIVING 8, DIED 1. Against that this archive holds SEVEN names — George William 1884, Ada May 1887, Fredrick James 1889, Agnes Kate 1892, MARTHA MAUD 1895, Florence Margaret H 1901, Ernest Albert C 1904 — of which six came from the GRO birth index searched on the mother's maiden name WALLIS. SO TWO ARE MISSING AND ONE OF THE NINE DIED BEFORE APRIL 1911. WHERE THEY WILL BE: the gaps are 1889-1892 and 1895-1901, and the second gap is six years wide in a family that otherwise produced a child every two or three. HOW TO FIND THEM: re-run the GRO birth index on mother's maiden name across ALL spellings, not only Wallis — Walls, Wallice, Wallace — because the missing two may simply be mis-indexed; then the Baldock burial register for a Castle infant. WHY IT MATTERS BEYOND COMPLETENESS: an unidentified son is another chance at a Castle male line, and the only Castle male line this archive has is Frederick James, who has now vanished twice. WORKED THE SAME AFTERNOON, AND THE METHOD FAILED IN AN INSTRUCTIVE WAY. The plan was to re-run the birth index on every spelling of WALLIS. THAT IS NOT THE PROBLEM. MARTHA MAUD'S OWN BIRTH ENTRY — Q3 1895, Hitchin, volume 3A page 561 — HAS NO MOTHER'S MAIDEN NAME IN THE INDEX AT ALL; THE FIELD IS BLANK. So the search that found six was never going to find nine, whatever it was spelt: it returns the children whose entries happen to carry the field and drops the others in silence, and the woman this archive descends from is one of the ones it drops. A FILTER ON A FIELD THAT IS SOMETIMES EMPTY IS A FILTER ON THE INDEXING, NOT ON THE FAMILY. THE DISTRICT SEARCH DOES NOT RESCUE IT EITHER: Castle births at Hitchin between 1884 and 1906 run to FOUR PAGES, because there are at least three unrelated Castle families in the town. WHAT IS LEFT, AND IT IS THE RIGHT INSTRUMENT ANYWAY: the BAPTISM REGISTER, which names both parents on every line. Martha Maud was baptised at Baldock on 22 September 1895, so the family used it. Findmypast's Hertfordshire Baptisms set has Castles at Baldock back to 1693 and would not take a year filter in the URL, so this needs the search form rather than a crafted address — or the Hertfordshire Archives register itself. THE TWO MISSING CHILDREN ARE STILL MISSING, and the gaps are still 1889-1892 and 1895-1901. WORKED TO ITS BOUND 17 SEPTEMBER, AND THE ROW IS STILL OPEN — but it is open for a known reason now. The plan was the BALDOCK BAPTISM REGISTER, which names both parents on every line. ALL 823 CASTLE BAPTISMS IN THE HERTFORDSHIRE SET WERE SORTED BY MOTHER'S FIRST NAME AND THE CAROLINE BLOCK READ. IT CONTAINS EXACTLY ONE ROW. Between the Barbara block and the Catharine block there is a single Caroline, and it is ADA MAY CASTLE, BAPTISED 10 APRIL 1887 AT WESTON. THE REASON IS A COVERAGE BOUND, NOT A SEARCH FAILURE: sorted by baptism year descending, the LATEST of 195 Castle baptisms at Baldock is 1879, and Caroline's children were born 1884 to 1904. THE INSTRUMENT ENDS FIVE YEARS BEFORE THE FAMILY STARTS. Martha Maud's own Baldock baptism of 22 September 1895 is not in the set either. SO THE TWO UNNAMED CHILDREN REMAIN UNNAMED, and the answer is at HERTFORDSHIRE ARCHIVES in the parish register itself, not in any index anybody can search from here. AND THEN IT CLOSED, the same afternoon, from the household rather than the register. THE 1901 CENSUS HAS THE WHOLE FAMILY: RG13 piece 1298, folio 49, page 26, SCHEDULE 192, NORTON END, NORTON, BALDOCK — George W Castle 36 a farm labourer, CAROLINE 36 born WESTON, and five children: George W 16, Ada 14, FREDERICK 11, Agnes K 9, and MARTHA 5. THEY HAD GONE BACK TO NORTON END, the hamlet where George William was himself raised by his GIDDINS grandparents in 1871 and 1881 after his mother died. He brought his own family home to it. THE NINTH CHILD IS SARAH ANN WALLIS, born 19 August 1882 at Maiden Street, Weston — TWO MONTHS BEFORE HER PARENTS MARRIED, AND REGISTERED UNDER HER MOTHER'S MAIDEN NAME. She is not a Castle birth in the index at all, which is exactly why no search for Castle births could ever have returned her. AND THE ONE WHO DIED IS NAMED AND DOCUMENTED: SIDNEY CHARLES CASTLE, born Q2 1898 at Hitchin and DIED Q4 1900 AT HITCHIN, AGED TWO. One entry in the birth index, one in the death index, no competitors. He was gone five months before the 1901 census, which is why no census ever showed him. NINE BORN, EIGHT LIVING, ONE DIED — AND THE ARITHMETIC NOW CLOSES EXACTLY: Sarah Ann 1882, George William 1885, Ada May 1887, Fredrick James 1889, Agnes Kate 1892, Martha Maud 1895, Sidney Charles 1898-1900, Florence 1901, Ernest 1904. THE SIX-INSTEAD-OF-NINE WAS THREE SEPARATE INDEXING FAILURES, not one: Sarah Ann is filed under WALLIS; Martha Maud's entry has a BLANK mother's maiden name; and Sidney Charles never came back on that search either. A FILTER ON A FIELD THAT IS SOMETIMES EMPTY IS A FILTER ON THE INDEXING, NOT ON THE FAMILY — and this row is the proof of it.

THE EXCLUSION IS NOW DOCUMENTED RATHER THAN ASSERTED, 17 September. Frederick James Castle, born 1889 at Hitchin to a mother nee Wallis, is Martha Maud's brother and THE ONLY CASTLE MALE LINE THAT COULD EVER CARRY A Y TEST. The obvious candidate death — FREDERICK JAMES CASTLE, Hitchin, 1975 — BELONGS TO SOMEBODY ELSE. The 1939 Register has that man at 17 ORCHARD ROAD, BALDOCK, BORN 10 APRIL 1887, 'Vertical Driller On Aircraft Work', married to Ella, with a third person in the household whose record is officially closed. Same name, same town, two years out. AND BALDOCK SUPPLIES A THIRD: the 1921 census has an entirely separate Castle household at 5 CEMETERY ROAD — head James Castle, nine children — with its own Frederick, 33 years 2 months, a hydraulic press operator at Letchworth. THREE MEN OF ONE NAME IN ONE SMALL TOWN INSIDE TWO YEARS, which is exactly why name-only identification at Baldock is worthless. WHAT IS LEFT OF OUR MAN: he is in no Hertfordshire census in 1911, in no census in 1921, in no 1939 Register entry, and in NO ENGLISH DEATH INDEX BETWEEN 1889 AND 2007 under any Fred-form of the name. FOUR THINGS COULD EXPLAIN THAT and they are testable: he died in the Great War, at 25 in 1914 (check Soldiers Died and the CWGC register); he emigrated, which this family demonstrably did; he died abroad; or he is the one of the nine who was dead by 1911, in which case the 1889 birth and an infant burial at Baldock should meet. THE FIRST IS THE ONE TO RUN, because a war death would also be a record naming his mother and her address. 17 September, while checking Frederick: THE 1901 CENSUS PRODUCES A THIRD BALDOCK FREDERICK AND HE IS NOT OURS EITHER. RG13 piece 1298 folio 48 page 23 schedule 175, NORTON END, BALDOCK: head JAMES CASTLE, 38, farm labourer; wife ANNIE; ten in the household; and FRED CASTLE, 13, farm labourer, born 1888. The same family is at MIDDLE ROW, BALDOCK in 1891 — RG12/1105 f.45 — with Fred aged 4 and a brother ARTHUR aged 2. SO THE BALDOCK FREDS ARE: ours b.1889 to George William and Caroline; the b. 10 April 1887 man of 17 Orchard Road who died in 1975; and this one, b.1888, son of James and Annie of Norton End. THREE UNRELATED CASTLE HOUSEHOLDS AND THREE BOYS OF ONE NAME INSIDE TWO YEARS. It also explains the 1901 census hit that looked promising: it is the third family, not the first. THE TRAIL PICKS UP AGAIN, 17 September, and it runs through a barracks. HE IS IN THE 1901 CENSUS AFTER ALL: FREDERICK CASTLE, SON, 11, at NORTON END with George W and Caroline — RG13/1298 f.49 p.26 sch 192. The earlier 1901 search that came back wrong had found the OTHER Norton End Castle family, three schedules away. AND IN 1911 THERE IS A BALDOCK-BORN FREDERICK CASTLE, AGED 21, SINGLE, AT MEEANEE BARRACKS, COLCHESTER — RG14 piece 10304, schedule 9999, Military Road and Mersea Road. HE WAS A SOLDIER, which is why he is in no Hertfordshire household in 1911. THE AGE IS THE DISCRIMINATOR AND IT IS THIN BUT REAL: 21 at 2 April 1911 means born between April 1889 and April 1890, which fits ours and EXCLUDES the Norton End Fred of the other family (23) and the Orchard Road man of 1887 (24, and he is at Baldock that night). IT IS NOT PROOF. The entry says FREDERICK, not Frederick James, and it rests on one census age and a birthplace. IT IS LABELLED PROBABLE AND NOT MORE. WHAT IT CHANGES: the 1921 census has ONE Baldock-born Frederick Castle in the whole of England and Wales and it is the b.1888 man of the other family, so ours is gone from the country by 1921. A regular soldier of 1911 who is not in Soldiers Died and not in the death index either went abroad and stayed, or died in a service Soldiers Died does not cover. NEXT: the regiment at Meeanee Barracks in 1911, then the medal rolls and the service papers for a Frederick Castle of that regiment. THE UNIT IS NAMED, and it took the original image rather than the transcript. The Findmypast transcript leaves the occupation column empty. THE IMAGE DOES NOT. Column 5 of the barracks return — 'UNIT, or ARM OF SERVICE' — reads SCOTTISH RIFLES, written once at the head of the sheet and carried down every line by ditto. That is THE CAMERONIANS (SCOTTISH RIFLES), and the surrounding pages of the same return confirm it: Malcolm Campbell, Duncan Sinclair, and men born at Glasgow, Lanarkshire, Dunbartonshire, Ayrshire, Roxburghshire and Kirkcudbright. A Hertfordshire farm labourer's son among them is unusual and is exactly what the regular army did. He is a PRIVATE, 21, single, born Herts Baldock, at MEEANEE BARRACKS, Colchester, on the night of 2 April 1911. AND IT IS STILL LABELLED PROBABLE. The entry says Frederick, not Frederick James, and the identification rests on one age and one birthplace — which is precisely the shape of the three mistakes this archive has made in the last week. WHAT IT UNLOCKS: a named regiment turns an impossible search into an ordinary one. The Cameronians' medal rolls, the medal index cards and the service papers can all be searched for a Frederick Castle, and a service record would carry his NEXT OF KIN — which would name George William Castle and Caroline and settle the whole thing. AND THE REGIMENT FOUND THE MAN, the same afternoon. BRITAIN, FIRST WORLD WAR CAMPAIGN MEDALS, TNA WO 329/1120: FRED JAMES CASTLE, SERGEANT, SERVICE NUMBER 10061, CAMERONIANS (SCOTTISH RIFLES), previous unit '2nd Sco Rif. 10061 Sgt.', awarded the BRITISH WAR MEDAL AND VICTORY MEDAL. HIS MEDAL INDEX CARD, TNA WO 372/4 (Discovery D1734236), gives the whole career on one line: SERGEANT THROUGHOUT, THREE NUMBERS — 10061 Scottish Rifles, 195568 MACHINE GUN CORPS, and 7815432 Scottish Rifles. THIS MORNING THE IDENTIFICATION RESTED ON ONE CENSUS AGE. IT NOW RESTS ON TWO INDEPENDENT MATCHES ARRIVED AT FROM OPPOSITE ENDS: the medal roll says FRED JAMES, which is exactly how the GRO registered him at Hitchin in 1889; and it says CAMERONIANS (SCOTTISH RIFLES), which is what the 1911 barracks return says in column 5 over Frederick Castle's line. AND THE THIRD NUMBER EXPLAINS EVERY ABSENCE THIS ROW WAS BUILT ON. 7815432 IS AN EIGHT-DIGIT NUMBER AND THE ARMY BEGAN ISSUING THOSE IN 1920. HE WAS STILL A SERGEANT AFTER THE WAR — which is why he is in no 1921 household, no 1939 Register and no English death index. HE WAS A CAREER SOLDIER, and career soldiers of the 1920s were mostly abroad. STILL NOT PROVED, AND THE FILE SAYS SO: the medal roll carries no date of birth and no birthplace, so the link is forenames plus regiment and not a document stating identity. A SERVICE RECORD WOULD STATE IT, because it names next of kin — and next of kin would be George William Castle and Caroline. ONE THING TO CHECK RATHER THAN BELIEVE: he has the War and Victory medals and NO 1914 OR 1914-15 STAR, although the 2nd Battalion went to France in November 1914. Either he was not with them, or he was on home or garrison service until 1916, or the Star roll has simply not been searched. It has not. NEXT: the service papers. A pre-1920 regular's papers are in WO 363/364 if they survived the 1940 fire; a man still serving in 1920 may have a record held to this day by the Ministry of Defence and obtainable by a next-of-kin application.

FOUND 17 SEPTEMBER while running the full-text row, and it is the more useful half of that morning. Searching the ordinary record index rather than the full-text one, surname WEAR, country SOUTH AFRICA, returns 16,156 rows across sets this archive has never touched: SOUTH AFRICA, CIVIL MARRIAGE RECORDS 1801-1974; SOUTH AFRICA, GAUTENG, JOHANNESBURG, CEMETERY RECORDS 1840-2019; SOUTH AFRICA, CHURCH OF THE PROVINCE OF SOUTH AFRICA (ANGLICAN) PARISH REGISTERS 1801-2004; SOUTH AFRICA, WESTERN CAPE, DECEASED ESTATE FILES 1951-2011; SOUTH AFRICA, CAPE, PROBATE RECORDS OF THE MASTER OF THE HIGH COURT 1822-1990; SOUTH AFRICA, PIETERMARITZBURG ESTATE FILES 1846-1950; SOUTH AFRICA, NATAL CIVIL DEATHS 1863-1955 and NATAL CIVIL MARRIAGES 1845-1955; KWAZULU NATAL VITAL RECORDS 1868-1976; CAPE CEMETERY RECORDS 1872-2011; and the Dutch Reformed and Netherdutch Reformed registers. WHAT IT ALREADY PROVED IN TWO MINUTES: BOTH OF LEONARD'S MARRIAGES ARE INDEXED THERE — Rosebank, 7 September 1940 to Phyllis Agnes Robertson, and Johannesburg, 30 October 1951 to Marjorie Eileen Dunlop — in South Africa, Civil Marriage Records. This archive found both the hard way, through court files and a register image. WHAT IT COULD ANSWER NEXT, and these are rows that have been open for a fortnight: the JOHANNESBURG CEMETERY set covers 1840-2019 and WEST PARK IS A JOHANNESBURG CEMETERY, so LEONARD'S BURIAL — and with it the 1973-or-1979 question — may simply be sitting in an index; the WESTERN CAPE DECEASED ESTATE FILES 1951-2011 is the set behind the two Cape Town estates already being asked for by letter; and the ANGLICAN PARISH REGISTERS would carry a Rosebank marriage if it was Anglican rather than Catholic. A DECOY TO WRITE DOWN NOW, BEFORE IT COSTS SOMEBODY A DAY: there is an entirely separate WEAR FAMILY IN DURBAN — Cecil Joseph Wear and Katie, with children baptised at Durban from 1913 to 1931, and Cecil Wentworth Wear who married Doreen Mary James in 1933. They are not ours and they will dominate any surname search. WHY THIS IS BLOCKED AND NOT RUNNING: THE FAMILYSEARCH SESSION EXPIRED MID-SEARCH and the site now asks for the password. I do not sign anybody in. David, when you next have a minute, sign in to FamilySearch in the browser and say so, and this goes from blocked to done in an afternoon. RUN AND CLOSED 17 SEPTEMBER, AND THE ROW UNDERSTATED IT. THERE ARE NOT ELEVEN SOUTH AFRICAN RECORD SETS. THERE ARE THIRTY-TWO. All 3,502 FamilySearch collections were enumerated through /service/search/hr/v2/collections and filtered; the list, with ids and counts, is in data/familysearch-south-africa.tsv. Then surname WEAR was matched EXACTLY - q.surname.exact=on, which drops the 16,156 fuzzy rows to 111 - and ALL 111 WERE READ, together with 2 for LUWINSKI and 24 for FIGUEIREDO. THE FINDINGS FILE FIRST SAID FIVE NEW RECORDS AND THE HONEST NUMBER IS TWO. Three of the five were already on this site, one of them far better than the index has it: the Germiston register of Ivone's first marriage has been READ, entry 27 of 1948 before magistrate A. J. Rossouw, and Frances's remarriage to Rolf Eisenhardt on 23 May 1958 has been published on Kurt's page for days. Written down because the error is the useful part: AN INDEX ENTRY AGREEING WITH A PAGE YOU HAVE ALREADY READ IS NOT A DISCOVERY. WHAT IS ACTUALLY NEW. (1) PHYLLIS AGNES WEAR MARRIED ROBERT SELWYN NORQUOY ON 14 JUNE 1952 AT MALVERN, JOHANNESBURG, aged 37 - ark 1:1:6ZHR-NR73. The archive had her second surname from a court index and no idea where it came from; WLD 790/1957 is Leonard suing a woman who had been remarried for five years, and she remarried seven and a half months after he did. (2) KATHLEEN EDITH WEAR AND JAMES CECIL FOOT, 26 MAY 1945 - the marriage was known from GenDatabase, which is a transcription. THE RECORD IS NEW AND SO IS WHAT IT IS: an ANGLICAN parish register, the first documentary evidence of this family's denomination in South Africa. (3) One further marriage is withheld because the bride may be living; it is in notes/, which is gitignored, and David has the decision. THE DECOY THE ROW PREDICTED WAS REAL: the Durban Wears are the majority of the 111. A SECOND DECOY IS NEW - LEONARD ANTHONY WEAR of the Western Cape, a different man with a different middle name, whose wife's estate file is indexed. AND THE NEGATIVE HELD: Leonard Aubrey Robert Wear appears in FamilySearch's ENTIRE South African holdings EXACTLY TWICE, both marriages, and he is under none of WARE, WEIR, WIER, WEARE, WEARNE or WEARS either.

ANSWERED 17 SEPTEMBER, and the answer is no. The Pix Road find opened one specific question: LOUISA SHENTON WAS HENRY ROBERT WEAR'S SISTER, so a descendant of hers would carry the WEAR half of Derrick's ancestry AND NOT THE CASTLE HALF — the independent test this archive has wanted since the segment reading, and the one thing that could separate the couple. No children were in the 1939 household, and the transcript confirms it held EXACTLY THREE PEOPLE with no officially closed records hidden in it. SO THE INDEXES WERE ASKED DIRECTLY. Births, surname SHENTON, mother's maiden name WEAR: ONE RESULT — STANLEY W SHENTON, Q1 1924, HITCHIN. Deaths, surname Shenton, born 1924: STANLEY W SHENTON, Q1 1924, HITCHIN. BORN AND DIED IN THE SAME QUARTER. His mother was forty-one. AND THE MOTHER'S-MAIDEN-NAME TRAP WAS AVOIDED THIS TIME, because it had just caught this archive over Martha Maud: the district was searched as well, surname SHENTON, HITCHIN, 1916 TO 1936, WITH NO MOTHER'S NAME AT ALL. ONE RESULT. THE SAME ONE. Shenton is a Staffordshire surname and there is exactly one Shenton birth in the whole Hitchin district across twenty-one years. THE MARRIAGE: WILLIAM SHENTON married a WEAR at Hitchin in Q3 1918, when Louisa was thirty-five. THE REST OF THE HOUSEHOLD, now traced end to end: WILLIAM SHENTON, born December 1876, 'Incapacitated' in 1939, DIED Q3 1953 at Hitchin. LOUISE S A SHENTON DIED Q4 1967 AT HITCHIN, AGED 84 — and the initials close it, because her birth is registered as LOUISA SARAH A WEAR. AND THAT BIRTH GIVES A PLACE NOBODY HAD: Q1 1883, KINGSTON, SURREY. Robert Wear and Emma Hawkins married at Southwark in 1881 and their son Henry Robert was born at St Olave Bermondsey in 1889 — IN BETWEEN THEY WERE AT KINGSTON. WHAT THIS COSTS: the Wear half can no longer be tested through Louisa. WHAT IS LEFT: Henry Robert's other siblings, and the two Hull half-siblings ALBERT J b.1895 and EDITH M b.1897, who carry Emma's DNA but not Robert Wear's — which is a different and still useful test, because a Hull cousin should match Derrick on the HAWKINS side ALONE.

WITHDRAWN 18 SEPTEMBER, THE DAY AFTER IT WAS WRITTEN, AND THE ERROR WAS MINE. I wrote here that a search returned NINE ELECTORAL REGISTER ENTRIES FOR WILLIAM WEAR, Surrey and London, 1885 to 1900, and called it the best unworked source for the Bermondsey Wears. I HAD NOT OPENED ONE. Opened, EVERY ROW IN THAT RECORD SET SAYS 'NO NAME ON RECORD' and carries a PDF icon. They are page images of volumes, not entries about people. THREE CONTROLS SETTLE IT. (1) Searching WILLIAM WEAR with no place filter returns 3,478 rows and the locations are DEVON and DURHAM — Weare in Devon and the River Wear in Durham. It is matching a place name, not a surname. (2) Searching JOHN SMITH returns the same kind of nameless rows in Lancashire and Warwickshire, which is not what a name index does. (3) Searching LUWINSKI returns ZERO, and a nonsense name returns zero too — so the fields are not inert, they simply match something that is not an elector. The 'Surrey & London' locations I quoted yesterday came from MY OWN PLACE FILTER, not from any name. WHAT I DID WRONG, AND IT IS THE THIRD TIME IN A WEEK: I read a RESULT COUNT as a FINDING. The Giddins matches, the two archives-index zeroes, and now this. A count is a property of a search box. WHAT SURVIVES, AND IT IS NOT NOTHING: the Bermondsey volumes for 1885 to 1900 ARE digitised and ARE browsable page by page, and an electoral register is still an annual address record covering the years that bracket both brothers' deaths in 1890. IT IS AN AFTERNOON OF READING IMAGES, NOT A SEARCH, and it is now filed as that. The caution already attached still stands: before 1918 the franchise is a property franchise, so lodgers and all women are absent and a silence means nothing about a life. || CLOSED 20 SEPTEMBER. The withdrawal was published, the replacement set (1910-1932, name-indexed) was found and worked in row 165, and its limit was established. There is nothing further to do on this row and it should stop appearing as work.

FOUND 17 SEPTEMBER while following the one baptism the register did yield. THE TREE THIS ARCHIVE HARVESTED FROM A MATCH HOLDER RECORDS 'Ada May Castle, died Jul 1969, Hitchin'. SHE MARRIED ON Q4 1909 AT HITCHIN, volume 3A page 1171, aged twenty-two. AND THE DEATH INDEX HAS NO SUCH ENTRY: Ada-anything Castle born 1884 to 1890, all England and Wales, returns sixteen rows and NOT ONE is at Hitchin and NOT ONE is in 1969. A WOMAN WHO MARRIED IN 1909 DID NOT DIE A CASTLE IN 1969. This is the CAROLINE JAMES trap in miniature — a tree recording a death under a maiden surname its owner had not used for sixty years — and it is the second time in two days that the same mechanism has hidden a woman from this archive. HER MARRIED SURNAME IS UNKNOWN AND IS NOT GUESSED HERE. Findmypast's MarriageFinder offers THOMAS REDMAN or SIDNEY SMITH; those are page-mates in an index and not a marriage, and this archive has now refused that particular temptation three times in one day. WHAT SETTLES IT: the certificate, 3A/1171, which names the husband and her father on one line and would confirm George William Castle at the same time. AND THE GENERAL LESSON, WHICH IS WORTH MORE THAN ADA MAY: every married daughter in that tree with a death recorded under CASTLE is now suspect. Agnes Kate married a Brown. Martha Maud married a Wear. Florence and the rest need checking the same way. SETTLED 17 SEPTEMBER, AND IT TOOK NO CERTIFICATE. THE 1911 CENSUS HAS HER UNDER HER MARRIED NAME: RG14 piece 7595, schedule 300, 27 CHURCHYARD, HITCHIN — ADA MAY SMITH, WIFE, 24, BORN BALDOCK HERTS, MARRIED 1909, YEARS MARRIED 2, ONE CHILD BORN AND ONE LIVING; head AMOS SMITH; her husband SIDNEY SMITH; their son also Sidney. THREE STRANDS ARRIVED AT SEPARATELY AND THEY MEET: the GRO marriage index (Ada May Castle, Q4 1909, Hitchin, 3A/1171), the census household, and MarriageFinder's page-mate suggestion of a SIDNEY SMITH. THE THIRD STRAND WAS WORTHLESS ON ITS OWN AND THIS ARCHIVE SAID SO THREE TIMES TODAY. It counts for something now only because the first two were found without it. HER DEATH REMAINS OPEN AND THE TREE'S VERSION OF IT STILL DOES NOT EXIST. The tree says 'died Jul 1969, Hitchin' under CASTLE; there is no such death under Castle at Hitchin in any year. WHAT THERE IS: ADA MAY SMITH, BORN 1887, DIED Q3 1969 — AT SPILSBY, LINCOLNSHIRE. Q3 IS JULY TO SEPTEMBER, SO THE TREE'S DATE IS RIGHT AND ITS SURNAME AND DISTRICT ARE WRONG, which is what happens when somebody finds a date and fills the rest in from assumption. BUT SMITH IS THE COMMONEST SURNAME IN ENGLAND and Lincolnshire is a long way from Hertfordshire. THIS ARCHIVE IS NOT ASSERTING THAT THE SPILSBY WOMAN IS HER. What is settled is the surname she died under, and it was not Castle. WORKED 20 SEPTEMBER AND DEFEATED BY THE SURNAME, WHICH IS WORTH MEASURING RATHER THAN SAYING. Ada May Castle married Sidney Smith, and SMITH is where she leaves the record. England and Wales death index, forename ADA MAY, born 1885-1889: THREE HUNDRED AND SIXTY-ONE CANDIDATES nationally. Narrowed to Hertfordshire with a wildcard on the forename: seventeen Ada Smiths and NOT ONE OF THEM CARRIES THE MIDDLE NAME MAY. The only Hitchin entry is Rose Ada Smith, 1889-1974, who is somebody else. THEN THE 1939 REGISTER, which prints exact dates of birth and would have keyed the death index in one step: TWO Ada Smiths in the whole county in her birth window, neither of them May. AND HER HUSBAND IS GONE TOO - Sidney Smith, born 1879-1889, at Hitchin in 1939: NIL. SO THE SMITHS LEFT HERTFORDSHIRE, and a national search on that surname and that forename is not a search. WHAT WOULD ACTUALLY WORK, and neither has been tried: probate, which indexes by exact date of death and address; or the 1921 census, which would give the household ten years after the 1911 entry at 27 Churchyard, Hitchin and might give a destination. 12 OCTOBER 2026 — SEARCHED UNDER THE MARRIED NAME AT LAST, AND THE TREE'S DATE FAILS AGAIN. England & Wales Deaths 1837-2007, forename ADA, surname SMITH, HITCHIN district, NO YEAR FILTER: six entries, dying 1912, 1952, 1953, 1955, 1960 and 1963, born 1870, 1879, 1880, 1892, 1899 and 1910. NOT ONE DIED IN 1969 AND NOT ONE WAS BORN IN 1887. This row explained the tree's “died Jul 1969, Hitchin” as the Caroline James trap, a death filed under a maiden surname. THE EXPLANATION WAS RIGHT ABOUT THE SURNAME AND IT DOES NOT RESCUE THE DATE: she is absent under BOTH names she is known to have borne, so the tree's date is not misfiled, it is unsupported. AND SHE IS NOT SEPARABLE NATIONALLY: Ada Smith born 1887, all England and Wales, returns 143 deaths, and a death index carries no husband, no address and no parents to tell them apart. SHE MAY ALSO HAVE REMARRIED AND DIED UNDER A THIRD SURNAME, in which case no search of SMITH will ever find her. THE YEAR FILTER WAS CONTROLLED IN THE SAME BREATH per row 108's rule: surname SMITH, year of birth 1887, no forename returns 12,232 in the same set, so the filter is live and the 143 is real. HER DEATH IS NOT FOUND and the next instrument is a certificate or a burial register, not an index. In data/castle-nine-children.tsv.

FOUND 17 SEPTEMBER, and it partly rescues the row withdrawn this morning. The set this archive tested and withdrew is ENGLAND & WALES, ELECTORAL REGISTERS 1832-1932, which returns page images and indexes no elector: every row says 'No name on record'. THERE IS A SECOND, DIFFERENT SET — ENGLAND & WALES, ELECTORAL REGISTERS 1910-1932 — AND IT CARRIES NAMES. A search for Castle in Hertfordshire returns rows like ARTHUR FREDERICK CASTLE 1929 and 1930, and FREDERICK CASTLE 1919 and 1920. WHAT IT CANNOT DO is the thing the withdrawn row wanted: it starts in 1910, and the question was about BERMONDSEY 1885 TO 1900, which brackets the two Wear brothers' deaths in 1890. For that, the images still have to be read by eye. WHAT IT CAN DO: place the BALDOCK CASTLES year by year from 1919, which is the decade in which Caroline became Caroline James and moved to Letchworth, and in which Frederick vanishes. THE CAUTION FROM THIS MORNING STILL APPLIES IN A NEW FORM: before 1918 the franchise was a property franchise; from 1918 it widens but still excludes women under thirty until 1928. An absence is not a death. WORKED 17 SEPTEMBER, AND IT BOTH DELIVERS AND DISAPPOINTS. THE SET IS REAL AND IT IS NAME-INDEXED: surname CASTLE at BALDOCK returns 356 ENTRIES across 1919 to 1932, by name and by year and by season — spring and autumn registers separately. That is the annual address record the withdrawn row wanted, and it exists. AND A FREDERICK CASTLE IS ON THE BALDOCK ROLL EVERY YEAR FROM 1919. Thirty-one Fred-form entries in that polling district. BUT IT CANNOT TELL US WHICH FREDERICK, AND THAT IS THE DISAPPOINTMENT. The transcript was opened — BL/ER/1919/D20, British Library SPR.Mic.P.16/BL.H.57 — and it carries NAME, YEAR, SEASON, POLLING DISTRICT AND CONSTITUENCY AND NOTHING ELSE. NO ADDRESS. NO AGE. Baldock had THREE Frederick Castles — ours born 1889, the 17 Orchard Road man born 10 April 1887, and the Cemetery Road man born about April 1888 — and an index of bare names cannot separate them. THE ADDRESS IS ON THE PAGE IMAGE AND NOT IN THE INDEX, so this is an afternoon of reading images if it is ever to answer. ONE THING IT DOES SAY, THOUGH, AND IT FITS: ERNEST ALBERT CYRIL CASTLE IS NOT ON THE BALDOCK ROLL AT ALL. He was twenty-one in 1925 and should have been, had he still lived there. He does not, WHICH AGREES WITH THE 1921 CENSUS PUTTING HIM AT 9 BOSCOMBE PLACE, LETCHWORTH, as James James's step-son.

OPENED AND IMMEDIATELY BOUNDED, 17 September. The service record is the document that would turn a strong identification into a proved one, because AN ATTESTATION FORM NAMES NEXT OF KIN — and next of kin would be GEORGE WILLIAM CASTLE and CAROLINE of Norton End. IT IS NOT AT KEW. All 78 Fred* Castle entries in WO 363 (the 'Burnt Documents') and WO 364 (pension claims) were read: NOT ONE is Scottish Rifles, Cameronians, or number 10061. AND THAT IS THE EXPECTED ANSWER, not a disappointment. Those two series hold men whose service ENDED BY ABOUT 1920; a man still serving kept a live record that never went to the Public Record Office. His medal card gives him 7815432, an eight-digit post-1920 number. THE SERIES CANNOT CONTAIN HIM BY DEFINITION. WHERE IT IS: with the MINISTRY OF DEFENCE, Army Personnel Centre, Glasgow. WHY THIS IS BLOCKED AND OWNED BY DAVID RATHER THAN BY ME: the MoD releases a service record to the NEXT OF KIN, or to anybody on production of a DEATH CERTIFICATE plus a fee, and THIS ARCHIVE HAS NO DEATH FOR HIM AT ALL. So the application needs either a death record this archive has not found, or a next-of-kin claim, and a next-of-kin claim is a statement about who somebody is that only the family can make. I am not filling in a form on anybody's behalf and I am not claiming a relationship for anybody. THE CHEAPER THING TO TRY FIRST, AND IT IS MINE: he is absent from the English death index because he was abroad. A soldier of the 1920s died, married and had children in ARMY REGISTERS — the British Army chaplains' returns and the consular and service registers — and those are indexed separately from the ordinary GRO series. That search has not been run. THE CHEAPER THING WAS TRIED FIRST, THE SAME AFTERNOON, AND IT CAME BACK EMPTY. BRITISH ARMED FORCES AND OVERSEAS DEATHS AND BURIALS, surname CASTLE forename FRED*: ten results, none of them Scottish Rifles and none of them 10061. The nearest by name is FREDERICK J CASTLE who died in 1917 — PRIVATE 47783, NORTHUMBERLAND FUSILIERS, Roclincourt Valley Cemetery, France. BRITISH ARMED FORCES AND OVERSEAS BANNS AND MARRIAGES, surname CASTLE: ALL FIFTY-NINE READ across three pages and NOT ONE IS A FRED. Those Castles marry at Malta, Kolar, Istanbul, Singapore, Italy, Aldershot and Johannesburg. AND THE CONTROL WAS RUN BEFORE THE NEGATIVE WAS BELIEVED: John Smith in the same marriage set returns 251 results from Lisbon in 1726 onward, so the set is real and was pointed at the right place. WHAT IS LEFT, AND IT IS NARROWER THAN IT WAS: he did not die in the war, he did not die on army service abroad under that name, and he did not marry in an army chapel abroad under it. He is in no English census after 1911, no 1939 Register and no English death index. HE LEFT AND DID NOT COME BACK, and the rest of his life is in a country this archive has not identified. The service record remains the one document that would say which.

RAISED 17 SEPTEMBER, AT DAVID'S PROMPTING, AND IT IS A GAP RATHER THAN A DISCOVERY. This archive has published RG10, RG11, RG12, RG13, RG14, RG15, BT 27, WO 329, WO 372 AND WO 392 — hundreds of references to records held at Kew — AND HAD NEVER ONCE OPENED THE CATALOGUE THOSE REFERENCES BELONG TO. Every one came second-hand through a commercial index's transcript. THE FIRST HOUR PRODUCED FOUR THINGS. (1) EVERY CITATION VERIFIES, and two verify BETTER than they were published: WO 392/15 is specifically 'SECTION 5: SOUTH AFRICA (UNION DEFENCE FORCES)', which is exactly the claim made about Leonard Wear, and WO 329/1120 adds the page range 312-622. (2) WO 372/4/50790 — THE EXACT ITEM REFERENCE for Fred James Castle's medal card, in Kew's own words: 'Scottish Rifles 10061 Serjeant. Scottish Rifles 7815432 Serjeant. Machine Gun Corps 195568 Serjeant.' Findmypast gave only the piece, and PRINTED THE DISCOVERY ID D1734236 ON ITS OWN TRANSCRIPT. This archive had that id for hours and never followed it. (3) DP/12/1/14 and DP/12/1/15, BALDOCK ST MARY THE VIRGIN, Hertfordshire Archives — registers of baptisms 1880-1895 and 1895-1910. The register this archive has been gesturing at all day is now an orderable reference. (4) LUWINSKI RETURNS ZERO across 2,500+ British archives, filed as a bounded negative with the reason a catalogue is not a name index. A REUSABLE TOOL WENT IN WITH IT: tools/tna.py, with a --verify mode that reads a reference back from the catalogue. WHY THIS IS RUNNING AND NOT DONE: the whole back catalogue of citations should be run through --verify, not just the nine tried today; the Hertfordshire registers for NORTON and WESTON have not been found in Discovery at any phrasing and may need HALS asked directly; and the WO 392 South African POW sections should be enumerated so the Leonard Wear identification cites the right volume for each of 1943, 1944 and 1945. PUBLISHED 17 September: Discovery is now named on /sources/ with what it verified and what it corrected. THE FULL SWEEP IS DONE, 17 September. Forty-four distinct archival references were extracted from the data files and the pages and each was read back from the catalogue. FORTY-TWO VERIFIED EXACTLY ON THE FIRST PASS. TWO RESOLVED AFTER ONE CORRECTION. NONE WAS WRONG, MISSING OR MISATTRIBUTED. AND THE TWO WERE THE SAME MISTAKE TWICE, AND IT IS NOT REALLY OURS: RG 15/7079 and RG 14/7576 both carry a LEADING ZERO in the piece number, which is how FINDMYPAST PRINTS THEM AND NOT HOW KEW DOES. Without the zero both resolve, and both describe LETCHWORTH (PART) in the Baldock sub-district — exactly what this site says they are. THE CATALOGUE ALSO CONFIRMED THE CONTENT AND NOT MERELY THE EXISTENCE: RG 11/574 is '5K St James Bermondsey' (Jamaica Road), RG 10/641 is '5J St James Bermondsey' (the father's house), HO 107/1560 is 'BERMONDSEY (Street Indexed)', RG 12/1120 is '2C Watford', RG 13/1312 is 'St Albans', RG 14/10304 is 'Colchester South-East' (the barracks), and the three BT 27 pieces are Southampton February-March 1937, London April-June 1939 and London August-October 1939 — Leonard's sailing, his father's and his mother's, in the right ports in the right months. ONE FALSE POSITIVE WAS IN THE EXTRACTION AND NOT IN THE SITE: the pattern pulled 'ED 26' out of a census citation, where it means ENUMERATION DISTRICT 26, and the catalogue matched a Board of Trade series of that name. The site never cited it; the script did.

RAISED 17 SEPTEMBER out of David's question about the Leibholz and Luwinski families. THE ENTIRE LINE LIVED IN BERLIN AND THIS ARCHIVE HAD NEVER LOOKED IN THE BERLIN DIRECTORY. The Zentral- und Landesbibliothek Berlin has digitised the Berliner Adressbücher FROM 1799 TO 1970, one volume a year, each 6,000 to 7,500 pages, listing householders by NAME, TRADE AND STREET. It is free, needs no account, and answers a JSON query at digital.zlb.de. tools/zlb.py now wraps it. 1,047 PAGE HITS FOR LEIBHOLZ. 23 FOR LUWINSKI. WHAT CAME OUT FIRST — A SURNAME LEAVING A CITY, MEASURED WITH A CONTROL. Leibholz page hits per annual volume: 24 in 1933, holding between 15 and 24 through the thirties, then 12 in 1940, FIVE in 1941, six in 1942, TWO IN 1943. AND THE CONTROL IS NOT AN IDLE ONE: THIMM — Johanna Thimm of Sauerbaum is Kurt's own great-grandmother, East Prussian and not Jewish — runs 66, 53, 47, 65, 72, 68, 45, 75, 69, 63 AND 75 IN 1943. As common in the last volume as in the first. The scanner did not get worse, the Fraktur did not get harder, the book did not get smaller. Bergmann is flat at around 700. Hirschfeld declines but far less steeply, which this archive cannot explain and does not pretend to. THE LIMITS ARE WRITTEN INTO THE FILE AND THEY MATTER MORE THAN THE NUMBER: a page hit is a page and not a person; twenty-four to two is a small base; and DEPORTATION, EMIGRATION AND ADMINISTRATIVE EXCLUSION FROM DIRECTORIES ALL LOOK IDENTICAL IN A COUNT. Jews were being struck out of German commercial listings from 1938 quite apart from being deported. THE COUNT CANNOT TELL THOSE APART. AND A LEAD: THE LUWINSKIS OF BERLIN ARE IN THE GARMENT TRADE. 1933 and 1936 give ALFRED LUWINSKI, BÜGLER — a garment presser — at N31 PUTBUSSER STRASSE 24 in Wedding; by 1943 an A. Luwinski is a Kraftwagenführer, a lorry driver. LIZZY WAS A DRESSMAKER AND MILLINER. Luwinski is not a common Berlin name — twenty-three pages in a hundred and seventy years of directories — and Joachim Luwinski and Johanna Thimm will have had children besides Auguste. A BROTHER IS THE OBVIOUS GUESS AND IT IS A GUESS, and Putbusser Straße 24 is a checkable address. NEXT, AND THE ORDER MATTERS: read the ALPHABETICAL NAME SECTION rather than the street section for 1924 to 1943 and find JACOB at Meraner Straße 2, Schöneberg, year by year — the year he stops appearing is the year the record of his life in that city stops; find AUGUSTE before 1924 under LUWINSKI and after it under LEIBHOLZ, and after the war under FRICKE; and follow the post-war Luwinskis of 1957 to 1970, because Lizzy lived until 1975. ONE HONEST DIFFICULTY: the OCR is Fraktur and the columns jam, exactly as the South African gazettes did. Counting works; reading a named line needs the page image. AND IT IS NOW ON THE SITE RATHER THAN IN A FILE, 17 September. The year-by-year table with its Thimm control is published on /leibholz/ under 'A surname leaving a city', with the three limits stated in the same panel — a page hit is not a person, the base is small, and deportation, emigration and administrative exclusion look identical in a count. The Luwinski garment-trade entries are on /east-prussia/, labelled a lead and not a finding. Both sources are now named on /sources/. WORKED PROPERLY 20 SEPTEMBER, AND THE FIRST FINDING IS THAT THE METHOD WAS WRONG. The page counts in this row - 24 in 1933 falling to 2 in 1943 - COUNTED FOUR SECTIONS AT ONCE. A Berlin volume has Teil I (householders and firms, one entry each), Teil II (the trades index, where one firm recurs), Teil III (institutions) and Teil IV (the street directory, where every tenant recurs). In 1933, twenty-four pages carried LEIBHOLZ AND EXACTLY ONE OF THEM WAS TEIL I. A count across all four measures the book as much as the family, and the Thimm control does not rescue it because it has the same problem. The name section has been read on its own for every volume 1920 to 1943; it is not consistently titled and is identified by position. ITEM 9 IS ANSWERED. 1933, page 1671, read straight off: "LuwinSki Alfred Bügler N31 Putbusser Str 24" - ALFRED LUWINSKI, A GARMENT PRESSER, A HEAD OF HOUSEHOLD. And that SEPARATES HIM FROM THE AROLSEN ALFRED LUWINSKI BORN 09.09.1917, who would have been sixteen: a boy of sixteen is not a householder and not a qualified presser. Not the same man, and the likeliest reading is father and son. He stays age-compatible with being Lizzy's brother and that is still not proved. LUWINSKI REACHES THE HOUSEHOLDER SECTION IN ONE YEAR OF THE WHOLE RUN. AND A THREAD NOBODY HAD PULLED. The Leibholz entries in Teil I are not scattered households, they are BUSINESSES, and they are in the clothing trade: Jaques Leibholz & Co., KONFEKTION, W8 Krausenstraße, 1924 to 1934; M. Leibholz, TRIKOTAGEN und WOLLWAREN, C2 Rosenstraße 17, 1929 to 1938; and SAMUEL LEIBHOLZ, "HÜTE EN GROS" - HATS, WHOLESALE. KURT LUWINSKI KEPT A HAT SHOP, THE PICCADILLY HAT, AT 81 ELOFF STREET, JOHANNESBURG. LIZZY WAS A DRESSMAKER AND MILLINER. The trade did not start in Johannesburg. Also in the run: William Leibholz, Generaldirektor and STADTRAT, 1920; Felix Leibholz, herring and anchovies; Louis Leibholz; Leo Leibholz & Co. AND HERE IS WHERE IT STOPS, DELIBERATELY. From 1939 the only thing the text layer returns under the name is the printer's cross-reference "Leibholz s. a. Leipholz" and a LEIBHOLD Wilhelm, who is a different surname. IT WOULD BE EASY AND WRONG TO PUBLISH THAT AS THE FAMILY VANISHING IN 1939: page 1719 of the 1939 volume comes back as ONE UNBROKEN BLOCK of seventeen thousand characters with the columns interleaved, its running head still says Leibholz, and whether an entry sits under it IS NOT LEGIBLE. ITEMS 7 AND 8 - Jacob at Meraner Straße 2 year by year, and Auguste under three surnames - ARE NOT ANSWERED for the same reason. WHAT WAS ESTABLISHED TOWARDS THEM: the Meraner Straße house block is page 5637 of 1936 and page 5715 of 1938, and in 1936 THE TENANT LIST FOR MERANER STRASSE 1 TO 5 HOLDS NO LEIBHOLZ - so if the 1939 census is right that Jacob, Auguste and Siegfried were at number 2, they arrived after 1936. Two individual Leibholz entries do surface in the street section after the firms stop: a PENSIONÄR in 1938 and a WIDOW in Schöneberg in 1940, neither legible enough to name. Full working in data/berlin-adressbuch-names.tsv. AND THEN CORRECTED AGAIN, THE SAME AFTERNOON, ON THE SAME PAGE. The reading above — that from 1939 the only thing left under the name is the cross-reference — IS WRONG, and the cause is a typesetting convention rather than the record. IN A GERMAN DIRECTORY ONLY THE FIRST ENTRY OF A RUN SPELLS THE SURNAME AND EVERY ENTRY AFTER IT BEGINS WITH A DASH. Searching for the string 'Leibholz' finds one line in ten. Followed properly, THE 1940 VOLUME HOLDS TEN LEIBHOLZ HOUSEHOLDERS (corrected 23 September from ELEVEN; the eleven is 1939): a master butcher, a police official, a widow, a pensioner, a clerk, an administrator, a bank official, the hat wholesaler, and two Wilhelms. So this row has now measured the wrong thing twice in one day — first by counting four sections of the book as one, then by counting one line of a run as the whole run. THE THIRD READING IS THE ONE TO TRUST AND IT IS IN data/berlin-adressbuch-names.tsv. WHAT CAME OUT OF IT: the identification of Samuel (row 177), the 1938 decree visible in the listing, AND ITEM 8 PARTLY — 'AUGUSTE, Witwe, Schöneberg, BADENSCHE STR 9' in 1940, with a conflict stated rather than smoothed: Jacob Leibholz was not deported until 26 February 1943, so a widow in 1940 is either a different Auguste, an error, or something about that marriage this archive does not know. 1942 AND 1943 ARE NOT COUNTED AT ALL, because the dash run breaks in the text layer and a short run cannot be told from a lost one without the image. CLOSED 20 SEPTEMBER ON THE FOURTH READING, FROM THE PAGE IMAGES. The text layer was exhausted and this row said so. THE IIIF IMAGES ARE NOT: digital.zlb.de serves every page at full resolution and the Fraktur is perfectly legible at three to four times magnification — the same technique that rebuilt the West Park cemetery registers. ITEM 7 IS ANSWERED. 'Jaques Kaufm Schönebg MERANER STR 2' is in the 1938 volume and the 1939 volume AND NO VOLUME AFTER — and item 7 asked exactly that, because 'the year he stops appearing is the year the record of his life in that city stops'. THE ANSWER IS 1939. He was deported on 26 February 1943; the directory loses him three volumes earlier. Bracketed on the other side: the 1936 street section lists the tenants of Meraner Straße 1 to 5 and no Leibholz is among them, so the household arrived between 1936 and 1938. ITEM 9 was already answered. ITEM 8 IS NOT, and is now precisely bounded: Auguste appears in ONE volume, 1940, as a WIDOW at Badensche Straße 9 in the same borough — and Jacob was alive until 1943, so the archive will not join them. She has never been looked for under LUWINSKI before 1924 or FRICKE after the war. AND THE COUNT THIS ROW SPENT THREE READINGS TRYING TO GET RIGHT. Householders: 17 in 1938, 11 in 1939, 10 in 1940, 9 in 1941, 6 in 1942, 4 in 1943. BUT THE NUMBER IS NOT THE FINDING; THE NAMES ARE. In 1940 Samuel is listed as 'Israel'. In 1941 two more men carry it and both have lost their trade in one year — a FLEISCHERMEISTER becomes an ARBEITER, a VERTRETER becomes a BAUARBEITER. IN 1942 EVERY NAME CARRYING 'ISRAEL' IS GONE. The four who survive to 1943 are exactly the four who never carried it. THE DIRECTORY MARKS THEM AND THEN LOSES THEM. What the count still cannot do is tell deportation from emigration from erasure — but it is now a list of named households with trades and street numbers, so each can be chased one at a time.

DONE 17 SEPTEMBER, at David's request for a copy of the decree to show. IT IS THE ZWEITE VERORDNUNG ZUR DURCHFÜHRUNG DES GESETZES ÜBER DIE ÄNDERUNG VON FAMILIENNAMEN UND VORNAMEN, 17 August 1938, promulgated the 18th — REICHSGESETZBLATT 1938, TEIL I, NR. 130, SEITE 1044, signed by Stuckart for the Interior Ministry and Gürtner for Justice. THE PAGE SCAN IS NOW HOSTED ON THE SITE at /rgbl-1938-1044.png, from Wikimedia Commons out of the Austrian National Library's digitised Reichsgesetzblatt. It is PUBLIC DOMAIN as a published German law, which is why it can be shown rather than linked. AND READING IT RATHER THAN PARAPHRASING IT MOVED THE ARGUMENT. THIS ARCHIVE'S STANDING OBJECTION TO KURT CARRYING 'ISRAEL' WAS THAT HE HAD LEFT IN 1933 AND WAS IN JOHANNESBURG BY 1939. § 2 (3) DISPOSES OF IT IN ITS OWN WORDS: 'Hat der Anzeigepflichtige seinen Wohnsitz oder gewöhnlichen Aufenthalt im Ausland, so ist die im Abs. 2 Satz 1 vorgesehene Anzeige an Stelle der Ortspolizeibehörde DEM ZUSTÄNDIGEN DEUTSCHEN KONSUL zu erstatten.' The decree was written with emigrants in mind, names the consulate as the office you report to, and routes births and marriages registered abroad to Standesamt I in Berlin. A MAN IN JOHANNESBURG IN 1939 WAS INSIDE ITS REACH AND THE MACHINERY FOR REACHING HIM IS SPECIFIED IN THE TEXT. AND § 1 (2) CLOSES THE OTHER DOOR: the decree does not apply to 'Juden, die eine fremde Staatsangehörigkeit besitzen'. KURT HELD NONE. This archive established separately, and for entirely different reasons, that he never naturalised — an alien for thirty-five years who died one in 1968. THE EXEMPTION THAT WOULD HAVE RELEASED HIM IS THE EXACT THING HE DID NOT HAVE. WHAT IT STILL DOES NOT PROVE, AND THE PANEL SAYS SO: it does not establish that Kurt was classed a JUDE rather than a MISCHLING, because the decree says only 'Juden' and leaves the definition to the Reich Citizenship Law — that is still the Geltungsjude question, and the Berlin Gemeinde membership registers are still what would settle it. Nor is there any document showing Kurt personally notified a consul. WHAT THERE IS: a law that reached South Africa, a man it was not exempt from, and the name on his records.

WORKED 17 SEPTEMBER. It had sat on the sources page under 'identified, not yet worked' since 14 September. THE INDEX IS VISIBLE WITHOUT CONSENTING TO ANYTHING and it holds 213 Leibholz person entries. CONFIRMED FROM A SECOND INDEPENDENT BODY OF RECORDS: JAKOB LEIBHOLZ, 08.08.1884, SCHUBIN — exactly the Gedenkbuch; he appears three times, once French-spelled as JACQUES, which suggests a displaced-persons or tracing document rather than a German one. OSKAR LEIBHOLZ, 13.05.1887, SCHUBIN, four entries. SIEGFRIED LEIBHOLZ, 05.08.1925 — Kurt's brother, to the day. TWO THINGS THAT ARE NEW RATHER THAN CONFIRMING. FIRST, THE COMPULSORY NAME IS ON THE FAMILY'S OWN RECORDS: 'LEIBHOLZ, ISRAEL OSKAR' and 'LEIBHOLZ, GEORG ISRAEL'. The 1938 decree published on /kurt/ today is no longer only a law quoted at these people — here are two of them carrying what it forced on them. SECOND, THERE IS A TRACING CASE FOR OSKAR: T/D no. 52.971, 'LEIBHOLZ, OSCAR born 13.05.1887', FOURTEEN DOCUMENTS. Somebody, after the war, asked the International Tracing Service what had become of him. WHO ASKED IS INSIDE THAT FILE, and the likeliest askers are a brother who got out or a widow who did not. A DISCREPANCY TO CARRY RATHER THAN RESOLVE: Arolsen gives GEORG as born 07.06.1885 where the Gedenkbuch gives 7 June 1884. Same day, same month, different year. AND SAMUEL IS THE ONE ABSENCE — nothing under 'Leibholz Samuel', although Yad Vashem and the Gedenkbuch both have him. Only twenty-five of the 213 entries were read, so HE MAY BE ON A PAGE NOT YET TURNED. Recorded as unfinished, not as a negative. AND THEN KURT. LUWINSKI, KURT, 11.05.1912 — his date of birth to the day — DocID 11214358, filed under '1 Incarceration Documents / 1.2 Miscellaneous / 1.2.1 DEPORTATIONS AND TRANSPORTS / Part 24: Folder MFN01335', reference 10000375 24, five hundred documents. KURT LEFT IN 1933 AND DIED IN JOHANNESBURG IN 1968. HE WAS NOT DEPORTED. The obvious readings are that he is named INSIDE a document about somebody else — a next of kin, an address, an emigrated son on a form about his father — or that a five-hundred-document folder is simply mixed. THIS ARCHIVE HAS BEEN WRONG THREE TIMES IN A FORTNIGHT BY READING A GOOD FIT AS A FINDING and it is not doing it again on this subject. WHY THIS ROW IS BLOCKED AND OWNED BY DAVID. Opening any of these documents requires ticking Arolsen's DECLARATION OF CONSENT — 'I hereby declare my consent to using sensitive personal data in accordance with applicable national laws and provisions.' THAT IS A DECLARATION A NAMED PERSON MAKES ABOUT THEIR OWN CONDUCT. I do not tick it on anybody's behalf; the rule was set at the start of this work and has not changed. Everything harvested above came from the index, which needs no consent. WHAT TO OPEN FIRST IF DAVID CONSENTS, IN ORDER: (1) DocID 11214358, because it is Kurt and nobody knows why he is there; (2) tracing case T/D 52.971 for Oskar, because it names whoever was looking for him; (3) the Jacques-spelled Jacob record, because a French rendering means a particular kind of file. ONE PRIVACY NOTE FOR THE SAME SEARCH: the Luwinski index also returns two post-war births that this archive's own register already omits as LIVING PEOPLE, and they are not named here or anywhere on the site. AND THEN IT GOT MUCH BIGGER, THE SAME AFTERNOON. A search for AUGUSTE returned somebody this archive did not know was in the persecution records at all: LEIBHOLZ, AUGUSTE, MAIDEN NAME LUWWINSKY, BORN 28.08.1893. THAT IS LIZZY, and the archive had her as 1893-1975 with no date. IT NOW HAS 28 AUGUST 1893. The maiden name is spelt LUWWINSKY with a DOUBLED W — a seventh spelling of this surname, German rather than South African, now added to /names/. THEN THE DOCUMENT IDS WERE COMPARED AND THEY ARE THE SAME ONE. Jacob b.08.08.1884, Auguste b.28.08.1893 and Kurt b.11.05.1912 ALL RESOLVE TO DocID 11214358, folder MFN01335, reference 10000375 24, under Deportations and Transports. THE FATHER WHO WAS DEPORTED, THE MOTHER WHO WAS NOT, AND THE SON WHO HAD BEEN GONE TEN YEARS, ALL THREE NAMED ON A SINGLE SHEET. THIS MORNING'S CHARITABLE GUESS — that Kurt was 'named inside a document about somebody else' — NOW HAS EVIDENCE INSTEAD OF GOODWILL: the somebody else is his father, and his mother is on it too. WHAT IS STILL INFERENCE AND MUST STAY LABELLED: THE SHARED DocID IS A HARD FACT FROM THE INDEX; WHAT KIND OF DOCUMENT IT IS, IS NOT. It could be a transport list, a property declaration, a Gestapo household record or a post-war reconstruction. Nobody here has read it and nobody here will tick a consent declaration on David's behalf to. IT IS THE ONLY WARTIME GERMAN DOCUMENT KNOWN TO CARRY KURT'S NAME, and if David consents to one thing in this archive, this is the one. UNBLOCKED AND OPENED 20 SEPTEMBER, David having ticked the consent declaration. DocID 11214358 IS A BERLIN POLICE HOUSEHOLD REGISTRATION CARD, printed form Vordruck Nr. 258, nationality 'Pr'. Three rows: LEIBHOLZ, JACQUES — 8.8.84 — SCHUBIN — religion 'mos.'; his wife AUGUSTE GEB. LUWWINSKY — 28.8.93 — SAUERBAUM; and 'Kind: SIEGFRIED — 5.8.25 — Bln-Schöneberg'. THE CARD SPELLS HIM JACQUES, which is the third independent source for the form of his name after the 1938 directory and the Mapping the Lives index, and it gives SAUERBAUM as Auguste's birthplace from a Berlin police file rather than a census card. THE RIGHT-HAND COLUMN IS A DATED LIST OF ADDRESSES running from 1913 into the 1930s, each paired with a landlord's name — Landsberger Str, Barbarossastr 17, Freisinger Str 15 from 1.6.33 — WHICH IS THE FAMILY'S MOVE HISTORY AND HAS NOT BEEN FULLY TRANSCRIBED; the scan is 976 pixels and the hand is Kurrent. The reverse lists the seven other indexed names — Schmidt, Puff, Heinicke, Herzog, Biere, Koch, Nicksch — WHO ARE LODGERS, not family. ONE HONEST GAP: KURT IS INDEXED TO THIS DocID AND IS NOT LEGIBLE AS A ROW ON THE CARD. AND THE ROW NAMED ITS OWN NEXT STEP, WHICH IS NOW ALSO UNBLOCKED: 'THERE IS A TRACING CASE FOR OSKAR: T/D no. 52.971, LEIBHOLZ, OSCAR born 13.05.1887, FOURTEEN DOCUMENTS. Somebody, after the war, asked the International Tracing Service what had become of him. WHO ASKED IS INSIDE THAT FILE.' THAT FILE CAN NOW BE OPENED. It is the single most promising unopened thing in Arolsen for this family, because a tracing enquiry names the enquirer and their relationship. See the new row.

WORKED 17 SEPTEMBER. This archive had used ONE Yad Vashem list — Jews murdered in Auschwitz — and had never searched the database itself. LEIBHOLZ returns 88 VICTIM RECORDS, 3 DATA-SUBMITTER RECORDS and 8 further, across the List of Murdered Jews from Germany (47), the LIST OF DEPORTATION FROM BERLIN (41), PAGES OF TESTIMONY (18) and the Theresienstadt inmate list (8). THE TRANSPORT. Jacob's record, 13520936, carries a field this archive has never had: 'DETAILS OF TRANSPORT: TRANSPORT 30 FROM BERLIN TO AUSCHWITZ BIRKENAU, EXTERMINATION CAMP, POLAND, ON 26/02/1943.' The date has been on these pages for a fortnight. THE TRANSPORT NUMBER HAS NOT. It also gives his permanent residence as BERLIN and his CITIZENSHIP AS GERMANY. AND A SECOND MEMORIAL BOOK NOBODY HERE KNEW THEY WERE STANDING ON. That deportation record is sourced NOT to the Bundesarchiv but to the GEDENKBUCH BERLINS DER JÜDISCHEN OPFER DES NATIONALSOZIALISMUS, Freie Universität Berlin, Edition Hentrich, Berlin 1995 — BERLIN'S OWN MEMORIAL BOOK, a different work from the national one cited on every page of this site, and the source of the transport number. IT SHOULD BE CONSULTED DIRECTLY for Auguste, for Siegfried, and for an address. OSKAR, record 13449776: Berlin to THERESIENSTADT 17/03/1943, Theresienstadt to AUSCHWITZ 16/10/1944, DATE OF DEATH 16/10/1944 — the same day the second transport left. A DISCREPANCY RESOLVED IN THE BUNDESARCHIV'S FAVOUR: Yad Vashem gives GEORG as born 1884, agreeing with the Gedenkbuch and DISAGREEING with Arolsen's 1885. Two against one, and the two are the older and better-sourced. LYDIA LEIBHOLZ appears on the LODZ GHETTO list, 1885 and 1886 in two records — which is where this archive already believed Georg's wife Lydia née Perl died, on the same day he did. AND A BERLIN LEIBHOLZ FAMILY THIS ARCHIVE HAS NEVER ACCOUNTED FOR: WILHELM 1890, ELSE 1889, RUTH 1923, LIESELOTTE 1923, WALTER 1922, CHANA 1938, HEINRICH 1890, ALBERT 1886, MAX 1877, HANS ISRAEL 1912 of Preussisch Friedland — most on the deportation list. WHETHER ANY IS A COUSIN IS UNKNOWN AND IS NOT GUESSED. But Sally Leibholz and Friederike Schmul had more sons than this archive has names for, and a Berlin Leibholz of the right age is where a fifth would show. A DECOY WRITTEN DOWN BEFORE IT COSTS SOMEBODY A DAY: 'KURT LEIBHOLZ, 1927, BERLIN, MURDERED'. THAT IS NOT OUR KURT. Ours is Kurt LUWINSKI, born 11 May 1912, who left in 1933 and died in Johannesburg in 1968. A Kurt of the same city fifteen years younger, murdered, is exactly the near-match this archive has been caught by three times in a fortnight. TWO OF THE 18 PAGES OF TESTIMONY HAVE NOW BEEN READ — SEE THE ROW BELOW. The remaining sixteen were not, and a Page names WHO SUBMITTED IT — a living relative writing down a dead one. The three data-submitter results (Heinrich Leibholz 1890, Mathilde Binder 1888, Sally Binder 14/01/1884) are not obviously ours but were not run down.

FOUND 17 SEPTEMBER by opening two of the eighteen Pages of Testimony. A PAGE OF TESTIMONY IS A FORM FILLED IN BY A PERSON WHO KNEW THE DEAD, and it names that person and their relationship. ELLI STRUZIK NÉE LEIBHOLZ, Yad Vashem 13689593, Page item 1819733: born 07/11/1915 at Berlin — which matches this archive's register to the day; FATHER GEORG; MOTHER LYDIA; MOTHER'S MAIDEN NAME PERL; married to HARRY; a SECRETARY; twenty-six years old; deported BERLIN to the ŁÓDŹ GHETTO on 01/11/1941 and ŁÓDŹ to CHEŁMNO on 13/05/1942. SUBMITTED BY MARGOT COX, COUSIN. GEORG LAIBHOLZ, Page item 1684365: born 1883; married LIDIA née PERETZ; of Berlin; PROFESSION FACTORY OWNER. SUBMITTED BY IRMGARD SHTERN, NIECE. WHAT THESE TWO NAMES ARE NOT: KURT'S RELATIVES ON ANY EVIDENCE THIS ARCHIVE HOLDS. The position argued at length on /leibholz/ is that GEORG IS NOT JACOB'S BROTHER — Jacob born Schubin in Posen 8 August 1884, Georg born Hammerstein in West Prussia two months earlier, and two men cannot be brothers on those dates. Margot Cox and Irmgard Shtern belong to GEORG'S LINE. AND THAT IS PRECISELY WHY THEY ARE WORTH EVERYTHING. IRMGARD SHTERN IS GEORG'S NIECE, SO HER PARENT WAS GEORG'S SIBLING — AND GEORG'S PARENTS ARE THE ONE FACT THAT WOULD SETTLE WHETHER HIS LINE TOUCHES SALLY LEIBHOLZ AND FRIEDERIKE SCHMUL OF SCHUBIN AT ALL. A NIECE IS ONE GENERATION FROM THE ANSWER TO A QUESTION THIS ARCHIVE HAS HAD OPEN SINCE DAY ONE. THREE FACTS CONFIRMED THAT WERE ASSERTIONS: Georg's wife WAS Lydia née Perl, and it is on her daughter's Page in a cousin's hand; Elli's death is no longer 'after 13 May 1942' but 13 MAY 1942 AT CHEŁMNO; and Georg was a FACTORY OWNER, which this archive did not know. THREE DISCREPANCIES CARRIED RATHER THAN RESOLVED: Georg's birth year is now 1884 (Bundesarchiv and the Yad Vashem list), 1885 (Arolsen) and 1883 (this Page) — three sources, three years. Lydia's maiden name is PERL on one Page and PERETZ on the other. And Georg's place of death is AUSCHWITZ on his Page but CHEŁMNO via Łódź on the deportation lists; A PAGE OF TESTIMONY IS WRITTEN FROM MEMORY YEARS LATER BY SOMEBODY WHO WAS NOT THERE, so the deportation list wins and this archive keeps Chełmno. NEXT: read the other sixteen Pages, because one of them may be for Jacob, Samuel or Oskar and may be signed by a Leibholz; and search Yad Vashem by SUBMITTER for SHTERN and for COX, because a person who filled in one Page usually filled in several. WORKED 20 SEPTEMBER, AND THE WHOLE SURNAME WAS SWEPT RATHER THAN TWO PAGES. 99 results in the Shoah Names Database for LEIBHOLZ: 88 about victims, across seven sources — List of Murdered Jews from Germany 47, List of Deportation from Berlin 41, PAGE OF TESTIMONY 18, Theresienstadt inmates 8, List of Persecuted Persons 5, Lodz ghetto inmates 3, Jews Murdered in Auschwitz 2. AND THE REASON TO DO IT TODAY RATHER THAN ANY OTHER DAY is that the 1940 Berliner Adressbuch had just yielded TEN LEIBHOLZ HOUSEHOLDERS (corrected 23 September from ELEVEN), so for the first time the two lists can be laid beside each other. AT LEAST FOUR OF THE TEN ARE IN THE DATABASE: Samuel Israel (identified, by matching street and house number), Wilhelm born 28.12.1890 (murdered, deported from Berlin, Theresienstadt), Albert born 1886 of Berlin, and Heinrich born 1890 of Berlin. THE DATABASE'S OWN PROFESSION FACET AGREES WITH THE DIRECTORY: it lists a BUTCHER, and the directory has two Leibholz Fleischermeister. A city directory and a register of the murdered, describing the same handful of tradespeople. AND A NEW LEAD ON THE GEORG QUESTION. HANS ISRAEL LEIBHOLZ, BORN 1912 AT PREUSSISCH FRIEDLAND, murdered at Auschwitz. Preussisch Friedland is in KREIS SCHLOCHAU — the same district as HAMMERSTEIN, where Georg was born, and nowhere near Schubin, where Jacob, Samuel and Oskar were born. This does not make Georg a brother. IT MAKES HIM A FAMILY: a second, West Prussian Leibholz line, of which Hans is the first member found besides Georg and Lydia, and of which Irmgard Shtern — who called herself Georg's niece — is a survivor. THE BOUND, STATED FIRST RATHER THAN LAST: three of the four Berlin matches are NAME AND CITY ONLY. Berlin held four million people, this archive has twice published a person built out of two, and a name in two lists is a correspondence until an address or a date of birth joins them. Only Samuel has that. In data/yadvashem-leibholz.tsv. TWO MORE OPENED 20 SEPTEMBER, CHOSEN RATHER THAN TAKEN IN ORDER. The eighteen Pages of Testimony were filtered and the two picked were the ones the Berlin directory work had already identified a household for. WILLI LEIBHOLZ, item 1067433: 'He was a BUTCHER. Prior to WWII he lived in Berlin.' Born 1893, married, spouse ERNA, PLACE OF DEATH AUSCHWITZ 1943. ERNA LEIBHOLZ, item 1070804: born Berlin, permanent residence NEUKÖLLN, spouse WILLY, place of death AUSCHWITZ. AND THE SAME PERSON WROTE BOTH: EVA KHAVA [P.] HAMMER OLSHER, NIECE. She is a HAMMER, and Mapping the Lives gives Erna as ERNA NÉE HAMMER — so the survivor who recorded this household came through the wife's family, not the husband's. WHAT IT CLOSES: the Berliner Adressbuch has 'Wilhelm, FLEISCHERMEISTER, Neukölln, LENAUSTRASSE 25' in 1940, 1941 and 1942 and nothing in 1943. Yad Vashem gives the trade, the city, the wife and Auschwitz in 1943. FOUR INDEPENDENT SOURCES NOW DESCRIBE ONE MAN AND THEY AGREE — a directory, the 1939 census database, the name records and a form written from memory by a woman who survived. AND THERE ARE PHOTOGRAPHS: his record lists four documents, two of them of type PHOTOS. ONE INCONSISTENCY KEPT RATHER THAN SMOOTHED: the Page gives his age as 45 against a birth of 1893 and a death in 1943, which does not add up; Pages of Testimony are written from memory and their age fields are the weakest thing on the form. AND THE BOUND: NONE OF THIS CONNECTS WILLI TO KURT'S FAMILY. Jacob, Samuel and Oskar were born at Schubin; nothing here says where Willi was born beyond 'Germany'. FOURTEEN OF THE EIGHTEEN PAGES REMAIN UNOPENED, including one for HEINZ LEIBHOLZ born 1922, who is the son at the same address. WORKED AGAIN, LATER ON 20 SEPTEMBER, BY SUBMITTER RATHER THAN BY VICTIM — AND THIS ROW'S CENTRAL ARGUMENT IS WITHDRAWN. The row said: Irmgard Shtern called herself GEORG'S NIECE, so her parent was Georg's sibling, so her other Pages would give GEORG'S PARENTS. She filed SEVENTEEN Pages and three of them settle it. LIDIA LAIBHOLZ b.1884 — father ELIUS, mother MINA. KLARA LEWINZON b.1906 — father YULIUS, mother MINA. EMA MENDELZON b.1892 — father YULIUS, mother MEINE. Elius and Yulius are one name through the Hebrew; so are Mina and Meine. THREE SISTERS, and Irmgard is niece to all three. SO GEORG IS THE UNCLE BY MARRIAGE — IRMGARD'S BLOOD IS LYDIA'S FAMILY, NOT THE LEIBHOLZES'. Her Pages cannot give Georg's parents and never could, and Georg's own Page (item 1684365) has no father and no mother on it and gives his birthplace as the single word "Germany". The open question is exactly where it was. WHAT THE SEVENTEEN DID GIVE. (a) LYDIA'S PARENTS AND SISTERS — a wife who was a name and a maiden name is now a daughter of Julius and Mina with at least two sisters. (b) GIDEON SHTRUZIK, BORN 1940, marital status "child", of Berlin, MURDERED, father HARRY — almost certainly Elli's son and Georg's grandson, although HIS MOTHER IS NOT NAMED ON THE PAGE. A child this archive did not have. AND THE OTHER SUBMITTER SEARCH FOUND A SURVIVOR. SUBMITTER LAST NAME = LEIBHOLZ returns three records, all filed by SOPHIE LEIBHOLZ through RELICO, the World Jewish Congress Relief Committee in Geneva, on questionnaires "COMPLETED BY SURVIVORS IN SHANGHAI, 1946". She was looking for her brother SALLY BINDER (b.14.01.1884, tailor, Motzstrasse 13, Neukölln — deported to Auschwitz 1 March 1943 and murdered), his wife MATHILDE, and her brother-in-law HEINRICH LEIBHOLZ (b.16.10.1890 Berlin, tax agent, Hornstrasse 19, wife AGNES) — THE CONTROL CASE, who survived and died in 1948, and whom she did not know was alive. A SECOND SOPHIE LEIBHOLZ, NÉE ROTHENBERG, BORN AT SCHLOCHAU 09.01.1887, was murdered at Auschwitz on the same 1 March 1943 transport and is a different woman; the two are kept apart deliberately. SCHLOCHAU IS THE DISTRICT THAT CONTAINS HAMMERSTEIN, where Sally Leibholz — Kurt's grandfather — was born. In data/yadvashem-leibholz.tsv. STILL OPEN AND NOW A ROW OF ITS OWN: the Relico Shanghai questionnaires as a collection, and Sophie's husband's forename, which is on none of the three forms.

FOUND 17 SEPTEMBER, and it is the oldest unsourced claim on the front page of this site finally getting a source. The chart has said 'SALLY LEIBHOLZ & FRIEDERIKE SCHMUL, Hammerstein; married at Szubin, 1874' since day one, ON THE FAMILY TREE'S WORD AND NOTHING ELSE. YAD VASHEM RECORD 5415605, SAMUEL ISRAEL LEIBHOLZ: born 28/06/1877 at SCHUBIN; FATHER'S FIRST NAME SALI; MOTHER'S FIRST NAME FRIDERICKE; MOTHER'S MAIDEN NAME SCHMUL; married to FRIEDA, which this archive did not know he had; died at AUSCHWITZ on 10/04/1943; CONTENT NOTE 'DOCUMENT NUMBER 18258/1943'; source THE AUSCHWITZ DEATH REGISTERS held by the State Museum Auschwitz-Birkenau. SO THE GREAT-GRANDPARENTS ARE NAMED ON A GERMAN DEATH REGISTER KEPT AT AUSCHWITZ, AGAINST A NUMBERED ENTRY, FOR A MAN BORN AT SCHUBIN IN 1877. The spellings are SALI and FRIDERICKE — what the clerk wrote. It is the first primary source this archive has ever held for that generation. AND A TEST WAS BUILT OUT OF IT AND RUN: Yad Vashem can be searched on MOTHER'S MAIDEN NAME. SCHMUL returns people from Berlin to Łódź to Komarów — it is a common name — AND EXACTLY ONE LEIBHOLZ, Samuel. NO BROTHER OF JACOB AND SAMUEL IS RECORDED ANYWHERE IN THAT DATABASE WITH THOSE PARENTS, AND THAT IS A LIMIT OF THE RECORDS RATHER THAN OF THE FAMILY: Jacob's and Oskar's entries come from DEPORTATION LISTS, which have no parents' fields at all. Only a Page of Testimony or a death register carries parents, and of the four brothers only Samuel has one. WHICH ALSO SETTLES THE SHAPE OF THE GEORG QUESTION: he will never be excluded or included by a parents' field in Yad Vashem, because his surviving records do not have one either. IRMGARD SHTERN, HIS NIECE, REMAINS THE ROUTE.

NEW 17 SEPTEMBER, and it began as a mistake worth keeping. The FamilySearch index gives Kurt's third wife as FRANCES HOSTOWSKY, BERNER, SEFTON, LUWINSKI, and her 1958 remarriage to Rolf Eisenhardt was written up here as a discovery before Kurt's own page was checked - WHERE IT HAS BEEN PUBLISHED FOR DAYS. So the row is not what it started as. WHAT IS ACTUALLY WORTH DOING: Frances is the only person in this story whose whole adult life is legible in one line of an index - born Hostowsky at Vienna, married Berner, married Sefton, married Kurt on 29 February 1956, married Eisenhardt on 23 May 1958 - AND THE ARCHIVE HAS NEVER LOOKED FOR THE FIRST TWO. Berner and Sefton are two marriages and probably a widowhood, all in South Africa, all in the civil marriage index that has now been mapped. One search each. SHE IS NOT BLOOD AND THE ARCHIVE SHOULD SAY SO PLAINLY: this is about establishing whether the 1956 register's 'widow' is true, because that is the field Kurt's page leans on. || BLOCKED AGAIN 20 SEPTEMBER: THE FAMILYSEARCH SESSION HAS SIGNED OUT for the second time in four days. The search page now says 'Create a free account or sign in to see all available results' and the persona API returns 0 results for a query that returned 111 on 17 September — a zero that means signed out, not absent, which is exactly the kind of false negative this archive keeps catching. I do not sign anybody in. One sign-in in the browser unblocks this row and row 174 together. RUN 20 SEPTEMBER, the sign-in restored. BERNER AND SEFTON ARE NOT IN THE SOUTH AFRICAN MARRIAGE INDEX. Frances's two marriages before Kurt do not appear under either surname, which fits the 1956 register giving her birthplace as VIENNA — she probably married them in Europe, not here. WHAT DID COME OUT: the Western Cape Archives Records version of her 1958 marriage to Rolf Eisenhardt GIVES A PLACE THE CIVIL MARRIAGE INDEX DOES NOT — WYNBERG, CAPE PROVINCE. So after Kurt she left Johannesburg for the Cape. Her age is given as 43 in both versions. The row closes: her whole traceable South African life is three entries, and the two earlier husbands are a European problem.

NEW 17 SEPTEMBER. Kurt's page carries '1957 - files for divorce again, Kurt Luwinski versus Frances, formerly Sefton, formerly Berner, born Hostowsky' AND data/litigation.tsv HAS NO REFERENCE FOR IT. Eight court files are tabulated there and that action is not one of them. It must exist: Frances married Rolf Eisenhardt on 23 May 1958 and Kurt did not die until 1968, so the marriage was ended by a court and not by a death. FIND THE NUMBER. It is a Witwatersrand Local Division file of 1957 or 1958, in the same series as WLD 1950/1948, WLD 497/1951, WLD 2368/1951 and WLD 2542/1957. NAAIRS, when it is back, under LUWINSKI, LUWINSKY and LUWENSKI - the last of which hid WLD 2545/1950 for weeks. AND NOTE THE NEIGHBOUR ALREADY IN THE TABLE: WLD 2542/1957, Standard Bank versus K. Luwinski, sits in the same window. Whatever 1957 was for him, it was happening in more than one court. || BLOCKED 20 SEPTEMBER: NAAIRS has stopped answering altogether — not 503 as on 17 September but no response at all, portal and search service alike, four days running. See row 179. This row cannot move until it comes back, and marking it 'next' overstated what this archive can do. 11 OCTOBER — THE BLOCKER IS GONE. NAAIRS IS BACK AT A NEW ADDRESS UNDER A NEW NAME: https://www.nationalarchives.gov.za/search/, the “National Archival Information Search System”, free, no account, all twenty repositories, with a POST API at /search-api/api/search. The old entry point is still dead, so the service did not come back where this archive was watching. THIS ROW IS NO LONGER WAITING ON AN INSTRUMENT. The instrument's own measurements and the three negatives re-run through it are in data/naairs-is-back-2026-10-11.tsv. 11 OCTOBER — FOUND, AND NOT WHERE THIS ROW SAID TO LOOK. TAB WLD 1878/1956, “ILLIQUID CASE. RESTITUTION OF CONJUGAL RIGHTS. KURT LUWINSKI VERSUS FRANCES (FORMERLY SEFTON, FORMERLY BERNER, BORN HOSTOWSKY).” THE YEAR IS 1956, NOT 1957 — the year he married her — see correction 9500919. And the row's search plan would never have returned it: it said to try LUWINSKI, LUWINSKY and LUWENSKI, and the file came up under HOSTOWSKY, her maiden name. AND THE ROW NAMED ITS OWN TRAP AND DID NOT SEE IT: its last line points at “the neighbour already in the table, WLD 2542/1957, Standard Bank versus K. Luwinski, in the same window”. THAT NEIGHBOUR IS WHERE THE 1957 CAME FROM. In data/naairs-sweep-2026-10-11b.tsv.

NEW 17 SEPTEMBER, and it is the best future prospect this archive has for the question it exists to answer. "South Africa, Transvaal, Estate Files, 1950-1993", collection 2144007, is in its own words IMAGES OF COPIES OF DEATH NOTICES from probate estate files at the National Archives in Pretoria. A South African death notice NAMES THE DECEASED'S PARENTS, and the title spans Kurt in 1968, Henry Robert in 1969, both of Leonard's candidate years, and Martha Maud in 1986. ITS WAYPOINTS - read from /service/cds/recapi/collections/2144007/waypoints - SAY IT IS PUBLISHED FOR 1955 TO 1960 ONLY. Seven of them. The collection page says "This collection is being published as images become available". THE CONTROL SAYS THE REST EXISTS: the Orange Free State equivalent, collection 1407787, has fifty-seven waypoints running 1951 to 2006 and 3.9 million images. So the Transvaal gap is a publishing schedule and not a destroyed record. CHECK THE WAYPOINT COUNT PERIODICALLY. It is one HTTP request and the answer is a number.

NEW 20 SEPTEMBER out of the Berlin directory, and it is the first thing that connects Kurt's working life to his mother's family rather than to Johannesburg. SAMUEL LEIBHOLZ IS LISTED IN THE BERLINER ADRESSBUCH AS "HÜTE EN GROS" - HATS, WHOLESALE - in the householder-and-firms section from 1929 to 1938, at SW19 and then SW68. KURT LUWINSKI RAN THE PICCADILLY HAT AT 81 ELOFF STREET, JOHANNESBURG, FROM 1941 TO 1952, and gave his occupation as dress designer in 1933 and again in 1956. AUGUSTE 'LIZZY' LUWINSKI WAS A DRESSMAKER AND MILLINER. And the other Leibholz firms are Konfektion and Trikotagen - ready-made clothing and knitwear. THE WHOLE FAMILY IS IN THE RAG TRADE, ON BOTH SIDES OF THE EQUATOR. WHAT THIS IS NOT, YET: proof that Samuel the hat wholesaler is Samuel Israel Leibholz of the Auschwitz death register, document 18258/1943, whose father was Sali and mother Friderike née Schmul. It is the same uncommon surname in the same city in the same decade, which is a candidate and not a finding - the Bloom mariner rule. WHAT WOULD SETTLE IT: the Adressbuch gives the firm's street and number, and Samuel Israel Leibholz's address is recoverable from the 1939 census the way Jacob's was. TWO ADDRESSES, ONE COMPARISON. CLOSED THE SAME DAY IT WAS OPENED, AND IT IS A FINDING RATHER THAN A CANDIDATE. The row asked for two addresses and one comparison. THE ADDRESSES MATCH. The Berliner Adressbuch gives SAMUEL LEIBHOLZ, HÜTE EN GROS, at JERUSALEMER STRASSE 5/6 in 1933 and at KOMMANDANTENSTRASSE 29 from 1936 onward. The Bundesarchiv Gedenkbuch gives SAMUEL LEIBHOLZ, born Schubin 28.06.1877, residence Berlin (KREUZBERG), and this archive's own Leibholz page has carried the street for days: KOMMANDANTENSTR. 29, KREUZBERG. SAME STREET, SAME NUMBER. The wholesale hatter is Samuel Israel Leibholz of the Auschwitz death register. AND THE 1940 VOLUME LISTS HIM UNDER THE NAME THE LAW GAVE HIM. Not 'Samuel Leibholz' but '—Israel Hüte SW68 Kommandantenstr 29'. THE 1938 DECREE, WHICH THIS ARCHIVE ALREADY SHOWS AS A DOCUMENT ON KURT'S PAGE, OPERATING IN A TELEPHONE BOOK: the forename gone and the imposed one in its place. A second man on the same page is listed as 'A. Israel'. AND KURT LIVED AT THE SHOP. The atlas gives four Berlin addresses for Kurt and one of them is Kommandantenstr. 29. A SECOND of the four is Waghäuseler Str. 9/10 — which the 1936 and 1937 directories give as the address of OSKAR LEIBHOLZ, Kaufmann, Wilmersdorf, matching the Gedenkbuch's Wilmersdorf for Oskar. TWO OF KURT'S FOUR BERLIN ADDRESSES ARE LEIBHOLZ BUSINESS ADDRESSES. He grew up in the family's trade and took it to Eloff Street. ONE THING WITHDRAWN: the reading that Samuel 'moved postal district between 1936 and 1938' is an artefact — SW19 was renumbered SW68 and the street never changed. AND THEN PART OF IT WAS WITHDRAWN THE SAME DAY. The sentence 'AND KURT LIVED AT THE SHOP', and the claim that two of Kurt's four Berlin addresses are Leibholz business addresses, ARE CIRCULAR AND ARE WITHDRAWN. Those four addresses are not Kurt's — they are the four Leibholz households in the CENSUS OF 17 MAY 1939, carried on the atlas under a loosely worded pin. Saying the directory agrees with them is saying the same thing twice, and it went into a commit message before it was checked. WHAT SURVIVES, AND IT IS STILL THE ANSWER THIS ROW WANTED: the identification of Samuel is sound, because the NEW fact is not the address but the TRADE — the Gedenkbuch gives a man at Kommandantenstr. 29 and the directory gives a WHOLESALE HAT BUSINESS there. And the directory does the same for the other three: JACOB traded as JAQUES and the firm 'Jaques Leibholz & Co, KONFEKTION' printed directly beneath him is his; OSKAR was a Kaufmann; GEORG RAN A LAUNDRY at Michaelkirchstraße 26. Four men who had four addresses and no occupations now have occupations. THE HAT THREAD TO JOHANNESBURG IS STILL REAL AND IS NOW PROPERLY STATED: Kurt's mother was a dressmaker and milliner, her husband's family firm was ready-made clothing, his brother sold hats wholesale, and Kurt opened a hat shop in Eloff Street. THAT IS A FAMILY TRADE, NOT A SHARED DOORWAY.

Re-checked 20 September. On 17 September the portal at national.archives.gov.za/naairs.htm answered 200 and looked normal, and only the search service behind it - www.national.archsrch.gov.za/sm300cv/smws/sm300dl - returned 503 Service Unavailable. TODAY BOTH TIME OUT WITH NO RESPONSE AT ALL. That is four days. THE ROWS THIS STOPS, and they are not small ones: 137 (MOTA, 288 archive records), 140 (re-verify every NAAIRS citation now the endpoint has moved), 175 (the WLD number for Kurt versus Frances, 1957), and any new court file for anybody. THE ARCHIVE IS NOT GOING TO KEEP WRITING THIS DOWN AS A FINDING - a service being down is a fact about a service. It is recorded once, with dates, so that the citations already published can be defended as having been verified when the endpoint was up. || 27 SEPTEMBER 2026: THERE ARE TWO NAAIRS DATABASES, not one that moved. NARSSA's own page says the NEW fulltext database is INCOMPLETE while 8.3 million entries migrate, and that the OLD one gives "a more complete search". Probed today: the old sm300 service 503s, naairs.htm is 404. This row's re-verification job now has a second axis — not just re-reading citations through a working endpoint, but RECORDING WHICH DATABASE ANSWERED. See row 275 and data/naairs-two-databases.tsv. 11 OCTOBER — THE BLOCKER IS GONE AND THE ENDPOINT NOW HAS A NAME. NAAIRS returned at https://www.nationalarchives.gov.za/search/, the “NATIONAL ARCHIVAL INFORMATION SEARCH SYSTEM” — free, no account, all twenty repositories, with a JSON API at /search-api/api/search. The old Search Manager endpoint is still 503 and naairs.htm still 404, so the service came back somewhere else for the second time in a month. AND THE TWO-DATABASE QUESTION THIS ROW EXISTS FOR IS ANSWERED FROM NARSSA'S OWN PAGE: the migration warning is still published — 8.3 million entries, the “fulltext search” admittedly incomplete — and it ends “To search the complete database please click on: Search NAAIRS Database”, WHOSE LINK IS THAT PORTAL. So the “old complete database” this row went looking for as a separate address IS THIS ONE. Three of the six nils were re-run the same day under seventeen spellings with SMITH (47,936) and SCHMIDT (3,418) as live controls, AND ALL THREE HELD. In data/naairs-is-back-2026-10-11.tsv.

THE REMAINING HALF OF ITEM 8, carved off so it stops hiding inside a closed row. The Berlin directory has now been read properly under LEIBHOLZ and Auguste appears in exactly one volume, 1940, as a widow at Badensche Straße 9. SHE HAS NEVER BEEN LOOKED FOR UNDER HER OTHER NAMES. Before her marriage to Jacob she was AUGUSTE LUWINSKI — and the Luwinski surname reaches the Berlin householder section in only one year of the whole run, 1933, so that search is small and quick. After the war she was AUGUSTE FRICKE and she lived until 22 June 1975, which is inside the digitised run: the ZLB has the volumes to 1970. A post-war Berlin address for Kurt's mother would be the first record of her own later life this archive has ever had, and the volumes are free. ALSO WORTH ONE QUERY: whether the 1940 widow is her at all. Jacob was not deported until 26 February 1943 and the archive will not join a 1940 widow to a man alive in 1943 without something more. WORKED THE SAME DAY, AND HALF OF IT IS ANSWERED. FIRST, A CORRECTION TO THIS ROW'S OWN PREMISE: the digitised run under the collection this archive had been querying STOPS AT 1943, and 1944 to 1954 return nothing. The claim that ZLB has 1799 to 1970 in one series is wrong. BUT THE POST-WAR BOOKS ARE THERE UNDER OTHER IDENTIFIERS — 34117222 for the Namenteil volumes of 1955 to 1970, 15849358 for the telephone books to 1980 — AND THEY ARE SET IN ANTIQUA, so the text layer reads them perfectly and no page images are needed. FRICKE: ONE CANDIDATE. 1957 Namenteil page 518: 'Fricke Auguste Rentn Chlb 1 Kais-Friedr-Str 79' — AUGUSTE FRICKE, A PENSIONER, CHARLOTTENBURG, KAISER-FRIEDRICH-STRASSE 79. Lizzy was born 28 August 1893 and died 22 June 1975, so in 1957 she was sixty-four and 'Rentnerin' is exactly right. She is the only Auguste Fricke in the Berlin name sections of 1955 to 1980. IT IS A CANDIDATE AND NOT A FINDING: Fricke is a common North German surname and Auguste a common forename for that generation, and one name in one directory is how this archive got the Bloom mariner wrong. WHAT WOULD SETTLE IT is her death registration of 22 June 1975, which would carry an address to compare. AND THE POST-WAR LUWINSKIS, WHICH THE ROW DID NOT ASK FOR AND WHICH MATTER MORE. MANFRED LUWINSKI, Schlosser, Tegel, Egellsstraße 16, in 1957, 1959 and 1961 — the register holds Manfred Joachim Luwinski, born 14 November 1933. JOACHIM LUWINSKI, Ofensetzer, Berlin 65, Ramlerstraße, in 1967 — the register holds Joachim Luwinski, born 30 November 1940. AND KURT LUWINSKI, POSTFACHARBEITER, BERLIN 21 (MOABIT), STEPHANSTRASSE, in 1964, with a telephone in the 1970-71 book. THAT LAST ONE SOLVES A STANDING BUILD PROBLEM. Every check-evidence run reports that 'KURT LUWINSKI' matches two register entries and is therefore left unlinked — Kurt Israel of Johannesburg who died in 1968, and a Kurt who died in 1995. THE SECOND KURT IS NOW PLACED: Berlin, the post office, Moabit, a telephone in 1970. The two are separated by city, trade and decade instead of by a death date alone. STILL UNDONE: Auguste under LUWINSKI before 1924. The surname reaches the householder section in only two pre-war years, 1909 and 1933, and the 1909 page has not been opened — it would be her father's generation. THE LUWINSKI HALF IS NOW CLOSED, AND FOR A STRUCTURAL REASON RATHER THAN A SEARCH. Auguste married Jacob in 1924, so before that she was AUGUSTE LUWINSKI — but she was born in 1893 and UNMARRIED WOMEN ARE NOT HOUSEHOLDERS. The Berliner Adressbuch's Teil I lists heads of household and registered firms; a woman in her twenties living at home or in service does not appear in it. AND THE SURNAME CONFIRMS IT: across 1900 to 1943 LUWINSKI reaches the householder section IN ONE YEAR ONLY, 1933, with Alfred the presser and Oskar the master tailor. There is no Luwinski household in Berlin for her to be part of before 1924. The apparent 1909 exception was opened and is a mirage — the page runs LÖWENTHAL to LÖWISOHN and the search fuzzes LU to LO. SO THERE IS NOTHING TO FIND HERE AND THE ROW SHOULD STOP IMPLYING THERE IS. THE FRICKE HALF STANDS AS ONE CANDIDATE: Auguste Fricke, Rentnerin, Charlottenburg 1, Kaiser-Friedrich-Straße 79, in the 1957 Namenteil — the only Auguste Fricke in the Berlin name sections of 1955 to 1980, and the right age. WHAT WOULD SETTLE IT is her death registration of 22 June 1975, which would carry an address to compare with Kaiser-Friedrich-Straße 79. That is a German civil record and it is David's to order. THE FRICKE HALF SWEPT PROPERLY, 20 SEPTEMBER, WITH THE SOLR API FOUND IN THE TEIL IV WORK. A phrase search for "Fricke Auguste" across ALL FORTY POST-WAR VOLUMES — 27 Namenteil 1949-1970 and 13 telephone books 1969-1980 — returns THREE PAGES, AND THEY ARE TWO WOMEN. 1957 Namenteil p.518: "FRICKE AUGUSTE RENTN CHLB 1 KAIS-FRIEDR-STR 79", the candidate this row already had. 1963 street section p.850 and 1970 p.287: "FRICKE AUGUSTE PUTZFRAU" and "FRICKE AUGUSTE" — THE SAME BUILDING BOTH YEARS, among Fröhlich, Kaldenbach, Luckow and Stinka. A WORKING CLEANER IN 1963 IS NOT THE SEVENTY-YEAR-OLD PENSIONER OF 1957. Two different women, and the Charlottenburg pensioner appears in NO OTHER VOLUME OF THE RUN AND IN NO TELEPHONE BOOK. Four Fricke blocks were also opened directly — 1952, 1954, 1963, 1970 — and none holds an Auguste under her own name or on the dash. WHAT THE SHAPE MEANS IF THE 1957 WOMAN IS LIZZY: she was in Berlin in 1957 and gone by 1959 — AND SHE WAS STILL IN BERLIN TWO YEARS AFTER HER MOTHER DIED AT ZEUTERN ON 30 OCTOBER 1955. She did not go to Baden to nurse her mother; she went afterwards, and died at Kraichtal, the same town, in 1975. THE IDENTIFICATION IS STILL UNPROVED AND NOTHING HERE CHANGES THAT — one line, no age, no birth date, no maiden name. What the sweep buys is the difference between "there is a candidate in 1957" and "there is a candidate in 1957 AND IN NO OTHER YEAR". BOUND STATED RATHER THAN GLOSSED: the dash convention means a phrase search cannot find a dashed "—Auguste", and twenty-three volumes were not opened by hand. The row stays blocked on her 1975 Kraichtal death registration. In data/berlin-postwar-luwinski.tsv.

NEW 20 SEPTEMBER, and it is one page. Sweeping the Berliner Adressbuch householder section for LUWINSKI across 1900 to 1943 returns the name in EXACTLY TWO YEARS: 1909, on page 1662 of 'I. Einwohner Berlins und seiner Vororte', and 1933, on page 1671, which is Alfred the presser at Putbusser Straße 24. THE 1909 PAGE HAS NOT BEEN OPENED. Auguste 'Lizzy' Luwinski was born on 28 August 1893, so in 1909 she was sixteen and cannot be the householder — WHOEVER IS ON THAT PAGE IS OF HER PARENTS' GENERATION, and this archive has never had a name for them. Luwinski is rare enough in Berlin that a single householder in 1909 is worth the four minutes: the page image is at digital.zlb.de/viewer/api/v1/records/34115495_1909/files/images/00001662.png and the Fraktur reads cleanly at three to four times. ONE HONEST CAUTION BEFORE ANYBODY GETS EXCITED: a Luwinski in Berlin in 1909 is not automatically her father. The family was East Prussian and the name appears in Posen and West Prussia too. STRUCK WITHIN THE HOUR, AND THE PAGE WAS OPENED BEFORE ANYTHING WAS WRITTEN ABOUT HIM. THERE IS NO LUWINSKI ON PAGE 1662 OF THE 1909 VOLUME. The page runs 'LÖWENTHAL — 1639 — LÖWISOHN' and is full of a completely different set of families: Semu LOWINSKI of Barnimstraße 50, Dr Julius LOWINSKIS of Brunnenstraße 196, WILLY LOWINSKY Ingenieur G.m.b.H., R. LOWINSKY Nachfolger Dampf-Waschanstalt, and a page of Löwenthals and Löwisohns. THE ZLB SEARCH FUZZES LU- TO LO- AND Ö-, and this archive had written a whole work-list row about Lizzy's father's generation before looking. IT TOOK FOUR MINUTES TO CHECK AND IT WOULD HAVE TAKEN A DAY TO UNPICK. WHAT OPENING IT DID PRODUCE, because the same doubt then had to be turned on the finding that IS published: the 1933 entry was re-verified on its own page image, and page 1671 runs LUTZ to LUX, which is the right alphabetical place, and the line reads 'Luwinski Alfred Bügler N31 Putbusser Str 24'. IT HOLDS. AND THERE IS A SECOND LUWINSKI ON THAT LINE'S DASH WHICH THE TEXT LAYER HAD HIDDEN: '—Oskar Schneidermstr W35 Steglitzer Str 7' — OSKAR LUWINSKI, A MASTER TAILOR. See the new row.

NEW 20 SEPTEMBER, found by re-verifying a published finding rather than by looking for him. The 1933 Berliner Adressbuch householder page carries TWO Luwinskis, not one: ALFRED, Bügler (a garment presser), N31 Putbusser Straße 24 — already published — and immediately beneath him on the dash, 'OSKAR SCHNEIDERMSTR W35 STEGLITZER STR 7'. OSKAR LUWINSKI, A SCHNEIDERMEISTER — A MASTER TAILOR — with his own business in Schöneberg. HE WAS LOST TO THE DASH CONVENTION, the same fault that hid four Leibholz households this morning: the text layer spells the surname only on the first line of a run. Twice in one day. WHY HE MATTERS. Luwinski reaches the Berlin householder section in ONE YEAR of the whole 1900-1943 run, so there were two Luwinski households in the city and BOTH WERE IN THE GARMENT TRADE — a presser and a master tailor. AUGUSTE 'LIZZY' LUWINSKI WAS A DRESSMAKER AND MILLINER, and her son opened a hat shop in Johannesburg. WHAT TO DO WITH HIM: he is not in this archive's register under any spelling. Look for Oskar Luwinski in the 1939 census via Mapping the Lives, in Arolsen, and in the post-war volumes beside Manfred, Joachim and Kurt. AND STATE THE OBVIOUS CAUTION: a rare surname and a shared trade is a reason to look, not a relationship. THE FIRST THING THIS ROW ASKED FOR WAS DONE THE SAME HOUR, AND THE ANSWER IS NO. Mapping the Lives — the memorial database built on the 1939 census supplementary card, which is where this archive's four Leibholz addresses came from — returns TWO PEOPLE for the surname LUWINSKI IN ITS ENTIRE HOLDING. One is Auguste née Luwinski at Meraner Straße 2, who is Lizzy. THE OTHER IS NEW AND IS NOT OSKAR. ANTON LUWINSKI, BORN 16 OCTOBER 1919, AT SCHINDEKOPSTRASSE 7, LEHRWERKSTATT OSTPREUSSISCHES BAUGEWERBE, KÖNIGSBERG (PR.) — a twenty-year-old at the East Prussian building trade's TRAINING WORKSHOP in Königsberg. THIS ARCHIVE HAS AN EAST PRUSSIA PAGE AND HAS NEVER HAD A TWENTIETH-CENTURY LUWINSKI IN EAST PRUSSIA. Note also that Arolsen holds an ALFRED LUWINSKI BORN 09.09.1917 — two young Luwinski men born two years apart, both in persecution databases. SO: OSKAR LUWINSKI THE MASTER TAILOR IS NOT IN THE DATABASE, AND NEITHER IS ALFRED THE PRESSER. For a database built on the census card that recorded Jewish descent, that absence is itself informative and this archive will put it no more strongly than that. WHAT IS LEFT OF THIS ROW: Oskar in Arolsen, and Oskar in the post-war volumes. AROLSEN CHECKED 20 SEPTEMBER, AND OSKAR IS NOT IN IT. Nine records under LUWINSKI in the whole Arolsen index and he is not one of them. Nor is he in Mapping the Lives. So the Schneidermeister of Steglitzer Straße 7 appears in ONE SOURCE — a single line in the 1933 Berliner Adressbuch — and in no record of Nazi persecution at all. THAT IS CONSISTENT WITH A MAN WHO WAS NOT PERSECUTED, and it is one absence in two archives, which is not proof of anything. WHAT IS LEFT AND CANNOT BE DONE FROM HERE: the post-war Berlin volumes have been read and hold Manfred, Joachim and Kurt but no Oskar; the 1939 census would have a card for him and Mapping the Lives does not index it, which means the card either does not survive or was never flagged. The row is parked rather than closed, because a second directory year would cost four minutes if anyone wants it. AND THE POST-WAR HALF OF THIS ROW IS NOW DONE TOO, THOUGH NOT FOR OSKAR. The whole post-war Berlin run was searched by surname for the first time. LUWINSKI RETURNS TEN PAGES IN EIGHT VOLUMES and THREE MEN, all three already in the tree and all three dead: MANFRED, Schlosser then MODELLSCHLOSSER, Tegel, EGELLSSTRASSE 16, in 1957, 1959, 1961, 1963, 1964, 1967 and 1970 — and MODELLSCHLOSSER IS MODEL FITTER, the exact trade on the 1965 ICEM emigration roll, which also names a MARQUARDT, and the tree holds Helga Erna Marquardt née Luwinski. Three sources, one man. JOACHIM, OFENSETZER — a stove-fitter — at RAMLERSTRASSE 15, Wedding, in 1967 and 1970: a trade and an address the tree does not have for him at all. KURT, Postfacharbeiter at Stephanstraße 7 in 1964 and Peter-Vischer-Straße 25 in 1967, and in the 1970-71 TELEPHONE BOOK as "LUWINSKI KURT 4 12 25 12 · BEAMTER · 52 WALDSTR. 60". NOT DERRICK'S FATHER, who died in Johannesburg in 1968 — this is the tree's other Kurt, Leo's son, recorded as dying in 1995 with no birth date, place or trade. A correspondence and not an identification. OSKAR LUWINSKI, THE MASTER TAILOR OF 1933, IS IN NONE OF THE FORTY POST-WAR VOLUMES. He does not reappear after the war under Luwinski, and LUWINSKY returns nothing in the run at all. The row stays parked on one source for him, and the sweep that would have unparked it has now been run and come back empty.

NEW 20 SEPTEMBER, out of a search that was looking for somebody else. Mapping the Lives holds exactly two Luwinskis: Lizzy, and ANTON LUWINSKI, BORN 16 OCTOBER 1919, recorded at SCHINDEKOPSTRASSE 7, LEHRWERKSTATT OSTPREUSSISCHES BAUGEWERBE, KÖNIGSBERG — the training workshop of the East Prussian building trade. WHY IT MATTERS: the Luwinskis are an East Prussian family and this archive's East Prussia page is built on nineteenth-century parish entries from Ermland and Masuria. ANTON IS THE FIRST TWENTIETH-CENTURY LUWINSKI THE ARCHIVE HAS FOUND IN EAST PRUSSIA, and he is in a persecution database, at an address that is an institution rather than a home. WHAT TO DO: read his full record for the source it came from and for any parents; check Arolsen for him; and set him beside ALFRED LUWINSKI BORN 09.09.1917, whom Arolsen holds — two young men born two years apart. THE CAUTION: Luwinski is rare but not unique, Königsberg is four hundred miles from Berlin, and NOTHING HERE CONNECTS HIM TO LIZZY except a surname and a province. That is a reason to look, not a relationship. RECORD OPENED THE SAME HOUR, AND IT CORRECTS THIS ROW'S OWN PREMISE. ANTON LUWINSKI was born at DEUTHEN, KREIS ALLENSTEIN — in the ERMLAND, the Catholic corner of East Prussia, and about twenty miles from SAUERBAUM, KREIS RÖSSEL, WHERE LIZZY WAS BORN. His card reads "Race": NNNN and "Basis of persecution: UNKNOWN". SO HE WAS NOT PERSECUTED, and this row's sentence calling Mapping the Lives "a persecution database" and drawing an inference from his being in it IS WRONG: the database indexes surviving 1939 census supplementary cards, and presence in it is not evidence of anything except that a card survives. WHAT IS LEFT, AND IT IS BETTER THAN WHAT WAS CLAIMED: two Luwinskis from two Ermland villages twenty miles apart, one of them Kurt's mother. Still a reason to look and still not a relationship. AROLSEN CHECKED 20 SEPTEMBER, AND ANTON IS NOT IN IT EITHER. Which adds nothing, and the reason it adds nothing is worth writing down: his own 1939 card already reads "Race: NNNN" and "Basis of persecution: unknown", so a persecution archive SHOULD NOT hold him, and it does not. The two sources agree and neither takes him further. WHAT HE STILL IS: a Luwinski born at DEUTHEN, KREIS ALLENSTEIN, twenty miles from Lizzy's Sauerbaum, both in the Ermland — the first twentieth-century Luwinski this archive has placed in East Prussia. WHAT WOULD TAKE HIM FURTHER is the same thing that would take Lizzy further: the Kreis Allenstein and Kreis Rössel parish registers, which are the undigitised films of the Groß Bössau row. HE IS NOW PART OF THAT ROW'S CASE RATHER THAN A SEPARATE LEAD.

OPENED 20 SEPTEMBER, and it is the single most informative document this archive has found about Kurt's mother. Mapping the Lives, from the Reichssippenamt census card of 17 May 1939, Bundesarchiv R 1509. (1) HER BIRTHPLACE. AUGUSTE LEIBHOLZ née LUWINSKI, born 28.08.1893 IN SAUERBAUM / RÖSSEL / OSTPREUSSEN — and this archive already knew that village, because JOHANNA THIMM OF SAUERBAUM is Kurt's great-grandmother and was used as the non-Jewish East Prussian CONTROL when counting Berlin directory pages. Lizzy was born where her grandmother came from. (2) "RACE": NNNN. The card recorded the ancestry of all four grandparents; N is nicht-jüdisch. FOUR N's. THE GERMAN STATE CLASSIFIED LIZZY AS NOT JEWISH, ON ITS OWN FORM, IN 1939. (3) "BASIS OF PERSECUTION: COLLECTIVE". The term for someone persecuted as part of a household rather than for their own recorded ancestry — A MISCHEHE. It is the standard reason Jacob was not taken until 26 February 1943, the Fabrikaktion. The category is on the record; the inference is not asserted. (4) AND IT GIVES JACOB A PLACE OF DEATH. The record's family block reads 'Leibholz, Jacob [Jakob], * 08.08.1884 in Szubin (Schubin), † before 08.05.1945 IN AUSCHWITZ'. The Bundesarchiv Gedenkbuch entry this archive holds says 'not documented' for both death date and death place. THE PLACE IS NOW NAMED. The date is not. AND THE QUESTION IT SHARPENS. Kurt was born in 1912, Lizzy married Jacob in 1924, so Kurt is her son by another man and Jacob was his stepfather. IF LIZZY WAS NNNN THEN KURT'S CLASSIFICATION AS A JEW — he carried the imposed name ISRAEL out of Germany in 1933 — CANNOT HAVE COME FROM HER. It came from the father whose surname he does not carry, and this archive has no name for that man. In data/mappingthelives-leibholz.tsv.

NEW 20 SEPTEMBER and it is the most precisely defined open task this archive has. The 1939 census card gives Lizzy's birthplace as SAUERBAUM, KREIS RÖSSEL, OSTPREUSSEN. The FamilySearch catalogue names the parish in its own words, record koha:15497: 'Parish register of baptisms and deaths for GROSS BÖSSAU, Ostpreußen, now BIESOWO, Olsztyn, Poland. Also includes Kunzkeim, SAUERBAUM, Willims, and Kleisak (Kreis Rössel).' The series runs 1690 to 1897 on NINE FILMS. EIGHT OF THE NINE ARE NOT DIGITISED. The two that matter both return 'Film number not found' in the viewer while signed in: FILM 7948905, TAUFEN 1793-1890, which would hold JOACHIM LUWINSKI'S 1865 BAPTISM AND NAME HIS PARENTS — a generation this archive does not have; and FILM 7769256, GEBURTEN (KOPIE) 1881-1958, which would hold LIZZY'S BIRTH OF 28 AUGUST 1893. Film 7948906 holds HEIRATEN 1700-1895, which would hold Joachim's marriage to Johanna Thimm. THE ONE FILM THAT IS ONLINE, 102682510, WAS OPENED AND CHECKED: 683 images, and at image 591 the camera photographed the film's own title board reading 'GROSS-BÖSSAU / KRS.RÖSSEL / 1798-1847 / ✝'. Right parish, wrong years — its baptisms stop in 1792. THREE ROUTES, IN ORDER OF COST: order 7948905 and 7769256 at a Family History Centre, where undigitised films can be viewed; write to the Archiwum Archidiecezji Warmińskiej at Olsztyn, which holds the Ermland originals; or re-check the two film numbers periodically, which is one request each. AND A WARNING FOR WHOEVER GOES: THE VIEWER FILES THIS CATHOLIC PARISH UNDER 'GERMANY, LUTHERAN BAPTISMS, MARRIAGES AND BURIALS'. The catalogue says Katholische Kirche and the Ermland was Catholic; the title board settles it. Anyone trusting the collection label would conclude these are the wrong registers and go home. In data/ermland-sauerbaum.tsv.

NEW 20 SEPTEMBER, found while checking Arolsen for somebody else. THE WHOLE SURNAME RETURNS NINE RECORDS AND SEVEN OF THEM MATCH THIS ARCHIVE'S OWN REGISTER BY EXACT DATE OF BIRTH. KURT IS ONE OF THEM. LUWINSKI, Kurt, 05/11/1912 — DocID 11214358, '1 Incarceration Documents / 1.2.1 DEPORTATIONS AND TRANSPORTS / Part 24: Folder MFN01335', five hundred documents. THAT IS THE SAME DocID THIS ARCHIVE ALREADY HELD FOR JACOB AND AUGUSTE, so the three of them are in one folder, and Kurt's index date of birth matches the register to the day. The archive knew the DocID; it did not know Kurt was indexed under his own name in it. THE OTHERS, IN FOUR DIFFERENT DOCUMENT CLASSES: FELIX (19.02.1923) in the Allied registration of foreigners and persecutees, DocID 69931515; MANFRED (14.11.1933) AND HELGA (01.05.1934) SHARING ONE DocID, 81709354, in '3.1.3 EMIGRATIONS — correspondence and nominal rolls'; ALFRED (09.09.1917) in post-war evaluations, DocID 130648701. Two further entries are a name-variant card and a surname with no forename. TWO MORE ARE PEOPLE FLAGGED LIVING IN THE FAMILY TREE AND ARE NOT NAMED — the build's own guard would stop them. WHY THIS IS BLOCKED AND OWNED BY DAVID. Every document behind these entries requires the consent declaration to be ticked, and I do not tick declarations on anybody's behalf. Name, date of birth, DocID and archival path are all Arolsen shows without it, and all of them are now recorded. ONE TICK BY DAVID OPENS FIVE DOCUMENTS, and the one that matters most is not Kurt's — it is 81709354, an EMIGRATION file naming two Luwinskis born a year apart, which is the only document in any archive that might say where the Berlin Luwinskis went. In data/arolsen-leibholz-luwinski.tsv. OPENED 20 SEPTEMBER, THE SAME DAY, AND IT IS THE BEST DOCUMENT OF THE WEEK. DocID 81709354 is an INTERGOVERNMENTAL COMMITTEE FOR EUROPEAN MIGRATION NOMINAL ROLL, headed MELBOURNE, numbered BGO 65/135, marked "AUSTRALIAN GOVERNMENT FINANCIALLY ASSISTED", stamped SEPTEMBER 1965, filed under "3.1.3 EMIGRATIONS — transport by airplane". Lines 85 to 88 are one family: MANFRED LUWINSKI, born 14.11.33, aged 32, of BERLIN, occupation MODELL FITTER; HELGA, born 1.05.34, aged 31, of Berlin; and TWO CHILDREN AGED TEN AND THREE WHO ARE FLAGGED LIVING AND ARE NOT NAMED. Forwarding address: c/o Mr H. MARQUARDT, HACKETT, CANBERRA ACT — and the family tree carries Manfred's wife as HELGA ERNA MARQUARDT Wagner. HER FAMILY WAS ALREADY IN AUSTRALIA AND BROUGHT THEM OUT. THE BERLIN DIRECTORY LINES UP EXACTLY: Manfred is a SCHLOSSER at Egellsstraße 16, Tegel, in 1957, 1959 and 1961 and vanishes after 1961. Schlosser and model fitter are the same trade. HE VANISHES BECAUSE HE LEFT THE COUNTRY. AND THE TREE ALREADY KNEW WHERE HE ENDED UP WITHOUT KNOWING HOW: Manfred Joachim Luwinski, died 20 SEPTEMBER 2012 AT BRISBANE. This document is the journey between Berlin and Brisbane. In data/luwinski-emigration-1965.tsv.

NEW 20 SEPTEMBER, and it is a fault in the family tree rather than a gap in the records. The MyHeritage export makes ALFRED LUWINSKI, BORN 9 SEPTEMBER 1917, the father of MANFRED born 14 November 1933 and JOACHIM born 30 November 1940 — and gives Alfred's wife as MARTHA SCHUENKE, WHO IT SAYS DIED IN 1930. SO THE FATHER WAS SIXTEEN AT THE FIRST BIRTH AND THE MOTHER HAD BEEN DEAD FOR THREE YEARS. It cannot stand as written. AND IT MEETS THE SAME PROBLEM FROM THE OTHER SIDE: the 1933 Berliner Adressbuch has ALFRED LUWINSKI, BÜGLER, A HEAD OF HOUSEHOLD at Putbusser Straße 24, which fits a man old enough to father a child that year and does not fit a boy of sixteen. THREE READINGS AND THE ARCHIVE SHOULD NOT PICK ONE: (a) the 1917 birth date belongs to a different Alfred and has been attached to the wrong man; (b) there are two Alfred Luwinskis, father and son, and the tree has collapsed them; (c) Martha Schuenke's 1930 death is the error. WHAT WOULD SETTLE IT: Manfred's own birth registration of 14 November 1933 at Berlin, which names both parents and their ages. Berlin Standesamt records of 1933 are not open, but Manfred emigrated to Australia in 1965 and DIED AT BRISBANE IN 2012 — an Australian death certificate names the deceased's parents, and Queensland registrations of that date are obtainable. THAT IS A CHEAPER ROUTE TO ALFRED THAN ANYTHING IN GERMANY, and it is in David's own state. ANSWERED WITHIN THE HOUR, FROM THE DOCUMENT THE TREE ITSELF CITES. Arolsen DocID 130648701 is a registration card of the Berlin 'Amt für die Erfassung der Kriegsopfer', and every field on it contradicts the tree. ALFRED LUWINSKI, born 9.9.17, Geburtsort a POLISH place; Beruf LANDWIRT; religion KATH.; nationality POLN.; FAMILIENSTAND: LEDIG — SINGLE ; INHAFTIERT SACHSENHAUSEN; 'gestorben 11.7.42 15:20, Herz- und Kreislaufschwäche', death certificate 1924/42. The reverse gives his home: 'wohnhaft PRZASNYSZ, PIŁSUDSKIEGO 117' — Masovia, north of Warsaw. SO HE IS A POLISH CATHOLIC FARM LABOURER WHO DIED AT SACHSENHAUSEN AT TWENTY-FOUR AND NEVER MARRIED. HE IS NOT THE FATHER OF ANYBODY BORN IN 1933 OR 1940. AND THE TREE CITES THIS EXACT URL as its source for Manfred's 1933 birth AND for Martha Schuenke's 1930 death. THE CARD CONTAINS NEITHER FACT AND NEITHER PERSON. Reading (b) of the three offered above is the right one: there are two Alfreds and the tree has collapsed them — or rather three, because the 1933 Berliner Adressbuch's ALFRED LUWINSKI, BÜGLER, PUTBUSSER STRASSE 24, a head of household, is a third man and is now THE ONLY CANDIDATE THIS ARCHIVE HAS for Manfred's actual father. WHAT FOLLOWS: MANFRED AND JOACHIM'S FATHER IS UNKNOWN. The Queensland death certificate suggested above is still the cheapest way to a name. In data/alfred-luwinski-sachsenhausen.tsv.

NOTICED 20 SEPTEMBER while reading the tree against the Arolsen documents, and it is an exclusion this archive made for a good reason and never revisited. Row 77 swept every Wear in the Johannesburg municipal cemetery registers and stated its bounds honestly, including: 'it does not cover the Chevrah Kadisha, which is a separate database — A WEAR WOULD HAVE NO BUSINESS IN A JEWISH CEMETERY but the exclusion is stated rather than assumed.' THAT IS RIGHT ABOUT THE WEARS AND WRONG ABOUT KURT. The family tree gives his burial as CHEVRA KADISHA, JOHANNESBURG. He was registered as a Jew in Germany, carried the imposed name Israel out of it, and the tree says he was buried by the Johannesburg Jewish burial society. THE CHEVRAH KADISHA MAINTAINS ITS OWN BURIAL DATABASE and it is not in FamilySearch's municipal collection. ONE ENQUIRY WOULD GIVE A DATE OF DEATH, A PLOT AND USUALLY THE NAME OF WHOEVER ARRANGED THE BURIAL — which is the question row 57 could never answer for his family. AND IT MAY SETTLE A DISCREPANCY THIS ARCHIVE IS CARRYING: the work list says Kurt died on 15 DECEMBER 1968 and the tree says 15 DECEMBER 1969, with an estate numbered 7584/1969. The day and month agree and the year does not. STRUCK THE SAME DAY, AND THE ROW WAS WRONG BEFORE IT WAS WRITTEN. THIS ARCHIVE HAS ALREADY CONSULTED THE JOHANNESBURG CHEVRAH KADISHA BURIAL REGISTER. It is the source of the FIRST entry in data/corrections.json and of the third: 'Kurt Israel Luwinski, died: was Dec 15 1969, now 15 DECEMBER 1968 — Johannesburg Chevrah Kadisha burial register: died 15.12.1968, buried 16.12.1968, WESTPARK BLOCK L GRAVE 211, Yahrzeit 24 Kislev 5729, which is itself 15 December 1968'; and 'Morris Isaac Salkow ... Westpark block U grave 2469'. SO EVERYTHING THIS ROW ASKED FOR IS ALREADY HELD: the date, the plot, and the Hebrew date cross-checking the civil one. IT ALSO SETTLES THE DISCREPANCY THE ROW RAISED, in the opposite direction to the one I assumed: the tree says 1969, the burial register says 1968, AND THE CORRECTION HAS BEEN IN THE FILE ALL ALONG. And one thing worth carrying away rather than losing with the row: KURT IS BURIED AT WEST PARK — block L, grave 211, in the Jewish section. THE SAME CEMETERY AS HENRY ROBERT AND MARTHA MAUD WEAR, who are in grave 56045, section E.C. Derrick's biological father's parents and the man who raised his mother's household lie in one ground. THAT IS THE THIRD TIME TODAY I have raised as new something this archive already had. The pattern is worth naming: I have been reading the work list and the page text and not corrections.json, which is where the hardest-won facts are.

NEW 20 SEPTEMBER. He was opened because a family tree said he was somebody's father, and he was not. HE IS RECORDED HERE ANYWAY, because the archive found him and dropping him because he is not useful would be the wrong instinct. ALFRED LUWINSKI, BORN 9 SEPTEMBER 1917 in Poland, a CATHOLIC FARM LABOURER, SINGLE, of PRZASNYSZ, PIŁSUDSKIEGO 117, in Masovia about a hundred kilometres from the East Prussian border. Imprisoned at SACHSENHAUSEN. DIED THERE ON 11 JULY 1942 AT TWENTY PAST THREE IN THE AFTERNOON; the card gives 'Herz- und Kreislaufschwäche', heart and circulatory failure, which is the standard camp formula and is not a cause of death. Death certificate 1924 of 1942. The Polish diplomatic mission and the I.R.O. were notified after the war and a later hand has dated the card 16.5.72 — somebody was still asking about him thirty years on. ONE THING WORTH NOTING AND NOT OVERREADING: Przasnysz is Masovian and the archive's other Luwinskis are East Prussian and Posen. The surname is thinly spread across all three, which is why it is traceable at all and why it keeps putting the wrong man in the right slot.

NEW 20 SEPTEMBER, after opening four Arolsen documents with David's consent and finding TWO MISATTRIBUTIONS IN THE FAMILY TREE, both of the same shape. (1) ALFRED LUWINSKI, born 9.9.1917: the tree makes him Manfred and Joachim's father and cites his Arolsen record. THE CARD SAYS POLISH, CATHOLIC, FARM LABOURER, LEDIG — SINGLE — SACHSENHAUSEN, DIED 11 JULY 1942 AGED TWENTY-FOUR, of Przasnysz in Masovia. He fathered nobody. (2) FELIX LUWINSKI, born 19.2.1923: the tree gives his birthplace as GRAFENAU, GERMANY and cites Arolsen. THE DOCUMENT GIVES NO BIRTHPLACE — it is a Form 11 return of the Landrat Grafenau dated 19 AUGUST 1946, an Allied listing of foreigners in the American Zone, and its nationality field reads RUSSLAND. THE PLACE OF REGISTRATION HAS BECOME THE PLACE OF BIRTH. THE RULE, AND THIS ARCHIVE SHOULD APPLY IT TO ITSELF BEFORE ANYONE ELSE. An Arolsen index entry gives a name, a date of birth and a DocID. EVERYTHING ELSE IS ON THE DOCUMENT. Until the document is opened, any place, relationship, occupation or fate attached to that entry is somebody's guess — and once it is written into a tree it gets cited back as though the archive had said it. FOUR DOCUMENTS OPENED, TWO MISATTRIBUTIONS. WHAT TO DO: the remaining Arolsen entries this archive has recorded WITHOUT opening the document — the 191 Leibholz index entries in data/arolsen-leibholz-luwinski.tsv — should be treated as pointers and nothing more until somebody opens them, and the file should say so. THE AUDIT WAS RUN, 20 SEPTEMBER, AND IT IS NOW A TOOL RATHER THAN A ONE-OFF: tools/audit-tree.py. 164 PEOPLE, 59 CITED FACTS, THREE HOSTS. (A) AROLSEN: SEVEN CITED FACTS AND ALL SEVEN POINT AT THE SAME DOCUMENT — DocID 130648701, Alfred Luwinski's Sachsenhausen card. ONE of the seven is supported by it, his own birth date. THE OTHER SIX ARE FALSE: two parents' 'birth of son', Alfred's fatherhood of Manfred, Martha Schuenke's 1930 death and her 'birth of son', and Manfred's 1933 birth. The card names no parents, no wife and no children and says LEDIG. (B) FAMILYSEARCH: FORTY-SIX CITED FACTS AND ONLY TWO ARE WRONG — 7 exact, 25 where the record is one to forty-five days after the event, 9 with no record date to check against. AND THE FIRST PASS OF THIS AUDIT SAID THIRTY-THREE WERE WRONG, WHICH WAS ITSELF THE ERROR: those 25 are PARISH REGISTERS, where the record is the BAPTISM or the BURIAL and the tree correctly holds the birth or the death. Stanilaus born 27 September 1810 and baptised 14 October; Valentin dead 12 July 1878 and buried on the 15th. A crude date comparison calls every one of those a mismatch. THE TWO THAT ARE REAL ARE THE SAME ERROR TWICE: Ivone's and Johannes de Witt's marriage of 2 FEBRUARY 1948 both cite the record of KURT AND IVONE'S MARRIAGE OF 19 JUNE 1948 — the wrong husband, 138 days out. The right record is already in hand from 17 September: ark 1:1:6ZCM-J38T, Germiston. (C) NAAIRS: six citations, NONE VERIFIABLE — they are URLs with session tokens that would not resolve even if the service were up, and it has not answered since 16 September. (D) INTERNAL IMPOSSIBILITIES: THREE IN 164 PEOPLE, found by arithmetic alone. Two are Martha Schuenke, dead in 1930 with sons born 1933 and 1940. THE THIRD IS NEW AND PAYS FOR THE AUDIT: JOHANNES ANTONIE OELOFSE DE WITT, born December 1923 at Johannesburg, has a SON OF THE SAME NAME born CIRCA 1923 AT PRETORIA, and both die at SASOLBURG. ONE MAN, SPLIT IN TWO — and the duplicate is carrying a fact the original lacks: DEATH 2 APRIL 1963 AT SASOLBURG. SO IVONE'S FIRST HUSBAND HAS A DATE OF DEATH, AND IT WAS HIDDEN INSIDE A DUPLICATE. The split also invented a phantom wife and a spurious half-sister relationship. AND THE RULE THAT COMES OUT OF IT. The FamilySearch citations are good BECAUSE EACH ONE CARRIES THE RECORD'S OWN DESCRIPTION IN ITS TEXT, so a wrong one can be caught without opening anything. The Arolsen citations are bare URLs with nothing to check against, WHICH IS EXACTLY WHY SIX WRONG ONES SURVIVED. A citation that cannot be falsified from its own text is not much of a citation. In data/tree-audit.tsv.

NEW 20 SEPTEMBER, out of the tree audit, and it is both a correction and a gain. The tree holds JOHANNES ANTONIE OELOFSE DE WITT twice: record 9500005, born December 1923 at Johannesburg, son of Frederik Petrus Johannes de Witt and Delina Johanna Oelofse who married at Burgersdorp on 3 October 1921, with a death place of SASOLBURG and NO DATE; and record 9500160, 'his son', of the SAME NAME, born CIRCA 1923 AT PRETORIA, DIED 2 APRIL 1963 AT SASOLBURG. A father aged nought, and both men dying in the same small Free State town. THEY ARE ONE MAN. SO IVONE'S FIRST HUSBAND DIED ON 2 APRIL 1963, AGED THIRTY-NINE, AT SASOLBURG — a date this archive did not have, sitting inside the duplicate. WHAT THE SPLIT ALSO CREATED: a phantom wife, 'Unknown de Witt' (9500161), who exists only to be the duplicate's mother, and a 'half sister' relationship for Delina de Witt that is an artefact rather than a fact about anybody. WHAT TO DO: this is the MyHeritage export and this archive's rule is that the charts draw the export as given — so the fix belongs in data/corrections.json as a was/now correction, not as an edit to the harvest. AND ONE CHECK BEFORE IT GOES IN: a South African death index or estate file for de Witt at Sasolburg in 1963 would confirm the date independently, and Sasolburg is in the Free State, so it is the Orange Free State estate files — a set FamilySearch has fully digitised, 1951 to 2006, on 57 waypoints. DONE 20 SEPTEMBER 2026, AND THE ESTATE FILES WERE NOT NEEDED. The row proposed browsing 57 waypoints of unindexed Orange Free State estate images. THE NATIONAL DEATH REGISTRATION INDEX COVERS 1953-1967 AND HAS HIM: "South Africa, Civil Death Registration, 1953-1967", Entry for JOHANNES ANTONIE OELOFSE DE WITT, 2 APRIL 1963, SASOLBURG, ORANGE FREE STATE, AGED 40, BORN 1923 AT PRETORIA (ark 1:1:DD9Y-PLT2). The duplicate's date is confirmed from outside the tree, and the duplicate's BIRTHPLACE is confirmed too — Pretoria, which is the half 9500005 disagreed with. AND THE SAME SEARCH FOUND A SECOND MARRIAGE NOBODY HAD: JOHANNES ANTONIE OELOFSE DEWITT and ELSIE MARIE GEYER, 25 FEBRUARY 1950, groom aged 27 (ark 1:1:6ZHR-NVYN). Ivone's first husband remarried two years after she left him and outlived that marriage by thirteen years. THE ONE THING IT UNSETTLES: three officials recorded his age — 25 in Feb 1948, 27 in Feb 1950, 40 in Apr 1963 — and the three ranges overlap at APRIL 1922 TO FEBRUARY 1923. The tree's "December 1923" is outside all three. Declared ages are not proof; three of them agreeing is recorded, not asserted. Written up in data/dewitt-confirmed.tsv, on the Ivone page, and into corrections 9500005 and 9500160. THE ESTATE FILE IS STILL THERE if anyone ever wants the de Witt line itself, which is not this family's.

CARVED OUT OF ROW 170 ON 20 SEPTEMBER, BECAUSE THE CONSENT THAT BLOCKED IT IS NOW TICKED AND IT DESERVES ITS OWN LINE. The Arolsen index holds a TRACING AND DOCUMENTATION CASE, T/D NO. 52.971, for 'LEIBHOLZ, OSCAR born 13.05.1887' — FOURTEEN DOCUMENTS. A T/D file is opened when a PERSON ASKS THE INTERNATIONAL TRACING SERVICE what became of somebody, and IT NAMES THE ENQUIRER AND THEIR RELATIONSHIP TO THE MISSING PERSON. Oskar was deported to Theresienstadt in March 1943 and murdered at Auschwitz on 16 October 1944; his wife Gertrud died on 27 February 1943. SO WHOEVER ASKED WAS NEITHER OF THEM, AND WAS ALIVE AFTER THE WAR AND KNEW HIM WELL ENOUGH TO ASK. This archive's whole difficulty with the Leibholz generation is that it has no survivor to anchor it: Margot Cox and Irmgard Shtern came from Pages of Testimony and are attached to GEORG's family, which the archive argues is a different family altogether. A NAMED ENQUIRER ATTACHED TO OSKAR — who IS Jacob's brother, both born at Schubin — WOULD BE THE FIRST SURVIVOR ON KURT'S OWN SIDE. Fourteen documents, and the consent is already given. WORKED 20 SEPTEMBER, AND THE ANSWER IS THAT THE CONSENT DOES NOT REACH IT. The topic page for 06030302.0.052.971 serves METADATA AND NO IMAGES — no document list, no DocIDs, no viewer, and the row expander opens nothing — and it carries its own explanation: 'UNDER CERTAIN CIRCUMSTANCES, PARTS OF THE FILE ARE NOT DISPLAYED FOR DATA PROTECTION REASONS. In such cases, please contact the Arolsen Archives.' THE DECLARATION OPENS DEPORTATION LISTS, REGISTRATION CARDS AND POST-WAR REGISTRATIONS AND STOPS AT THE CORRESPONDENCE, which is consistent with what a T/D file is. ONE THING THE METADATA DOES GIVE: creation date 1949 TO 2002. The enquiry was opened in 1949 and the file was still being added to fifty-three years later. SO THE ACTION IS A LETTER AND IT IS DRAFTED, NOT SENT: requests/29-SEND-THIS-to-arolsen.txt, which asks for the enquirer's name and relationship, accepts that part may stay closed, and asks whether the 2002 activity means a second person asked. AND THE HUNT PRODUCED SOMETHING ELSE ENTIRELY, WHICH IS IN THE NEW ROW.

FOUND 20 SEPTEMBER while failing to open T/D 52.971. A DIFFERENT AROLSEN SERIES IS SERVED WITHOUT RESTRICTION: 6.3.1.1, 'Searching for missing persons — TRACING INQUIRIES 1945-1946'. Searching LEIBHOLZ there returns 43 topics, including a 'Personal file of LEIBHOLZ, ROSA' — a forename this archive has never seen — and 'Personal file of LEIBHOLZ, WILLI and of further persons', five documents, WHICH WERE READ. IT IS A JEWISH AGENCY FOR PALESTINE SEARCH BUREAU FOR MISSING RELATIVES FORM, JERUSALEM, QUESTIONNAIRE NO. 2, DATED 23 MAY 1946, FILE 50764. THE PERSON SOUGHT: LEIBHOLZ, WILLI, aged 55, BORN AT STOLP IN POMMERN; last information MARCH 1943 FROM BERLIN; last address BERLIN-NEUKÖLLN, LENAUSTRASSE 25; occupation BUTCHER. With him: ERNA, 48, born HAMMER, wife; HEINZ, 23, son; AND RUTH, 22, DAUGHTER — A CHILD THIS ARCHIVE DID NOT HAVE. THE PERSON WHO ASKED: TABORI-HAMMER born NACHUM, RUTH-TRUDE, of BENEI-BRAQ, JEHOSHUASTRASSE, PALESTINE, born 9 SEPTEMBER 1899 AT HAMBURG, last address abroad BERLIN-MOABIT, TURMSTRASSE 59, relationship 'BROTHER IN LAW AND HIS FAMILY', signed in her own hand. The ITS answered her on 8 December 1947 through the Jewish Agency and closed the case on 22 December. AND IT SETTLES THE QUESTION I LEFT OPEN THIS MORNING. I wrote then that 'nothing here says where Willi was born beyond Germany'. STOLP IN POMMERN IS NOT SCHUBIN IN POSEN — two hundred and fifty miles and a different province — SO THE LENAUSTRASSE LEIBHOLZES ARE DOCUMENTED AS NOT KURT'S FAMILY rather than merely unproven. TWO HAMMERS SURVIVED AND BOTH WENT LOOKING: Ruth-Trude in 1946, and EVA KHAVA HAMMER OLSHER, niece, who wrote Willi's and Erna's Pages of Testimony later. BOTH COME THROUGH ERNA'S FAMILY, NOT WILLI'S. THE HONEST DISAPPOINTMENT: the row wanted a survivor on KURT'S side, and the file that would give one is the closed one.

DONE 20 SEPTEMBER 2026, all three, and the third one was the prize. (1) T/D 755.334 FOR HEINZ IS RESTRICTED. Tested, not assumed: reference 06030302.0.755.334, creation date 1959-1988, eleven documents, the same data-protection sentence Oskar's file carries, and NO DocID, NO IMAGE, NO DOCUMENT LIST. The consent declaration does not reach the T/D repository, and 755.333 ERNA and 755.335 RUTH stand beside it — three consecutive case numbers for the Lenaustraße household. (2) THE ROSA FILE IS ONE SHEET OF PAPER. DocID 86338329, a cross-reference slip in red: "AJDC-Berlin · Leibholz, Rosa · X-Reference = Silbermann, Heinz; ", Else; ", Inge". No date, no place, so it cannot be tied to this family or ruled out; what it gives is a route, because her papers were filed under SILBERMANN. (3) OSKAR'S OTHER RECORDS — AND THEY CARRY HIM OUT OF THERESIENSTADT. The Theresienstadt central card file (1.1.42.2, DocID 5055941) holds a Czech card: "Leibholz Oskar · 13.5.1887 · DEPORTACE NA VÝCHOD: Číslo Er 606, dne 16.X.1944 · (původní transportní číslo: 11025-I/90)". THE ORIGINAL TRANSPORT NUMBER IS THE BERLIN TRANSPORT'S OWN DESIGNATION — the Gestapo list of 17 March 1943 names itself "(I/90)" in its scope note. The two ends of the journey name each other. The AJDC card (1.2.1.2, DocID 11243228) gives SCHUBIN/BROMBERG in full, the deportation address BERLIN W 50 MARBURGERSTR. 13, the filing "49/30999 II — 4.gr.Alterstrsp. 17.3.43", and "Ehefrau: (Sa) GERTRUD geb. LEIBKE 13.7.94 Berlin". The Berlin registration card (1.2.1.5, DocID 11214364) puts Oskar and Gertrud on one form with CONSECUTIVE KENNKARTE NUMBERS, A 442 462 and A 442 461. ON "Verw.Abz." THE ROW'S INSTRUCTION STANDS AND IS NOT BROKEN: the abbreviation is still not expanded here. What was learned is what kind of column it sits in — the same column carries "Geltungsjude", "Ehepaar", "Pflegekind" and "Sohn von Nr. 69", so it is a column of exemption and relationship categories. That narrows it and does not settle it. Written up in data/oskar-leibholz-theresienstadt.tsv and data/arolsen-tracing-files.tsv. RE-TESTED LATER ON 20 SEPTEMBER AFTER DAVID TICKED THE CONSENT AGAIN, AND NOTHING CHANGED. T/D 52.971 still serves fourteen documents' worth of metadata and no DocID, no image, no list; T/D 755.334 the same with eleven. THE RESTRICTION IS ON THE SERIES, NOT ON THE USER — which is now tested twice rather than assumed. The declaration was never the obstacle; it is what opens the SERVED collections, and it has been open all day, for Oskar's four records, Jacob's transport list, Gertrud's Zählkarte, the Shanghai CLI card and the Willi file. HEINZ'S FILE HAS BEEN ADDED TO LETTER 29 rather than sent as a second letter, and the letter now says plainly that both were tested twice with the consent in place.

SUPERSEDED 22 SEPTEMBER 2026 — DO NOT SEND THIS ONE. The draft at requests/30-SEND-THIS-to-bydgoszcz.txt asked Bydgoszcz for the 1874 Szubin marriage because it names both sets of parents. IT WAS NEVER SENT, AND IT NO LONGER NEEDS TO BE: the archive answered the 14 September letter on 22 September by pointing at genealogiawarchiwach.pl, and the marriage was read at the image the same hour — akt nr 9 of 24 November 1874, USC Szubin Miasto sygn. 141 scan 10. It gives Sally Leibholz as born and resident at Hammerstein, aged twenty-seven, son of the Kaufmann JACOB LEIBHOLZ and WILHELMINE GEB. POLLNOW. THE DOCUMENT THIS ROW EXISTED TO OBTAIN IS NOW READ AND PUBLISHED. The letter's own fallback question — whether Bydgoszcz holds the Szubin Jewish community records — is the one thing in it still worth asking, and it has been lifted out into its own draft at requests/36, which is row 266. NEW 20 SEPTEMBER, out of David's question about the SCHMUL family, AND IT IS THE CHEAPEST UNWORKED THING ON THIS LIST because the letter is already sent and acknowledged. The front-page chart says SALLY LEIBHOLZ & FRIEDERIKE SCHMUL, 'Hammerstein · married Szubin 1874'. FRIEDERIKE IS NOT IN FAMILYSEARCH'S POSEN CHURCH BOOKS — all 33 exact SCHMUL records there were read and she is not among them — AND SHE WOULD NOT BE, because those are CHURCH books and a Jewish marriage is not in one. PRUSSIAN CIVIL REGISTRATION BEGAN ON 1 OCTOBER 1874, so a Szubin marriage that year is either a Jewish community register entry or a STANDESAMT SCHUBIN entry. EITHER WAY IT IS AT THE ARCHIWUM PAŃSTWOWE W BYDGOSZCZY — the same archive this archive wrote to on 15 September for Standesamt Schubin 1884, Jacob's birth, and which acknowledged with a warning that complex queries take over a month. ADD THE MARRIAGE TO THAT THREAD RATHER THAN OPENING A NEW ONE. WHY IT IS WORTH MORE THAN THE BIRTH CERTIFICATE: A PRUSSIAN MARRIAGE CERTIFICATE OF 1874 NAMES BOTH SETS OF PARENTS with ages and occupations. It would give FRIEDERIKE SCHMUL'S FATHER AND MOTHER and SALLY LEIBHOLZ'S — the generation ABOVE the top of the chart, which this archive has never reached on any line of the German side. ONE THING TO SAY IN THE LETTER: the chart says the couple were 'of Hammerstein' and married at Szubin, and Hammerstein is in WEST PRUSSIA while Szubin is in POSEN. If the 1874 record is not at Bydgoszcz it may be at the state archive holding Hammerstein, and correction 5 — which rests on Georg being born at Hammerstein and Jacob at Schubin — WOULD NEED RE-EXAMINING, because a family with both places in it is exactly what that correction assumes is impossible. DRAFTED THE SAME DAY: requests/30-SEND-THIS-to-bydgoszcz.txt, in Polish with an English translation beneath, in the house style used for Koszalin. NOT SENT. It is written as a REPLY ON THE EXISTING THREAD rather than a new enquiry, so it lands against the same file, and the note at the top says so — the archive has not recorded which address was used on 15 September and THE LETTER DOES NOT INVENT ONE; the sent mail has it. WHAT IT ASKS: the 1874 Szubin marriage of SALLY (Sali, Salomon) LEIBHOLZ, born at Czarne 1847, and FRIEDERIKE born SCHMUL. It raises the 1 OCTOBER 1874 boundary itself rather than letting them discover it, asks them to check BOTH the Standesamt books AND the Szubin Jewish community register, and says that even an answer naming the right series would be worth having. IT ALSO TELLS THEM WHICH TO DO FIRST IF THEY CAN ONLY DO ONE: the 1874 marriage, not the 1884 birth, because Jacob's parents are known and his grandparents are not. And it gives them the reason it has become urgent — KOSZALIN ANSWERED ON 16 SEPTEMBER that the 1884 Czarne books are not available and that they do not know what became of them. SALLY WAS BORN AT CZARNE IN 1847. If those books are gone, the Szubin marriage may be the only surviving document that names his parents.

NOTICED 20 SEPTEMBER while answering David's question about the Thimms, and it is a join this archive already had the pieces for. The East Prussia page records JOHANNA THIMM as dying on 30 OCTOBER 1955 AT ZEUTERN NEAR BRUCHSAL, and AUGUSTE 'LIZZY' LUWINSKI as dying on 22 JUNE 1975 AT KRAICHTAL. IT LISTS THEM AS TWO PLACES. ZEUTERN IS A DISTRICT OF KRAICHTAL. Mother and daughter died in the same small Baden-Württemberg town, twenty years apart — out of the Ermland, through Berlin, and into the west. THE PAGE SHOULD SAY SO IN ONE LINE. AND IT BEARS ON THE FRICKE CANDIDATE RAISED THIS MORNING: Johanna's own addresses include RICHARD-WAGNER-STRASSE 42, CHARLOTTENBURG, from 24 July 1932, and the 1957 directory candidate is 'Fricke Auguste, Rentnerin, Charlottenburg 1, Kaiser-Friedrich-Straße 79' — THE SAME DISTRICT. That does not confirm it, and Lizzy ending at Kraichtal where her mother died pulls the other way. WHAT WOULD SETTLE IT IS ONE RECORD: Lizzy's death registration of 22 June 1975 at Kraichtal, which is a Baden-Württemberg Standesamt entry and is David's to order. It would also confirm or break the Charlottenburg address. DONE 20 SEPTEMBER 2026. The East Prussia page now says it in three places: the two entries in the key-value list, and a panel. ZEUTERN IS ONE OF THE NINE VILLAGES AMALGAMATED INTO KRAICHTAL IN 1971-75, which is why the 1955 death is registered under the village and the 1975 death under the town; Johanna's last address, Besingstraße 110 at Zeutern, is a Kraichtal street. The same edit carried a second thing found the same day: DocID 11214358, opened at last, gives AUGUSTE'S BIRTHPLACE AS SAUERBAUM IN HER OWN HOUSEHOLD'S HAND — a second independent source for what had rested on the 1939 census card alone — and gives SIEGFRIED'S BIRTHPLACE AS BERLIN-SCHÖNEBERG, which nothing else did. THE FRICKE CANDIDATE IS NOT ADVANCED BY ANY OF THIS and stays where it was.

NOTICED 20 SEPTEMBER while drafting the Bydgoszcz follow-up, and it deserves stating because it cuts against something this archive has published. CORRECTION 5 in data/corrections.json disputes Georg Leibholz's brotherhood on exactly this ground: 'Born Hammerstein, West Prussia, 7 June 1884; Jacob born Schubin, Posen, 8 August 1884. Two months and two hundred kilometres apart.' THE ARGUMENT TREATS HAMMERSTEIN AS A PLACE THIS FAMILY HAS NOTHING TO DO WITH. IT IS NOT. The front-page chart says the great-grandparents were 'Hammerstein · married Szubin 1874', and the Koszalin letter of 15 September asked that archive for 'Standesamt Hammerstein 1884 — Georg; AND SALLY, 1847'. SO SALLY LEIBHOLZ, JACOB'S OWN FATHER, WAS BORN AT HAMMERSTEIN. The family had people in both towns. THAT DOES NOT MAKE GEORG A BROTHER — two sons of one mother cannot be born two months apart, and that arithmetic is untouched — BUT IT REMOVES THE GEOGRAPHY FROM THE ARGUMENT. A Leibholz born at Hammerstein in 1884 is no longer a stranger who happens to share a surname; he is a Leibholz born in the town Jacob's father came from, which is what a COUSIN looks like. Open question 5 asks 'Is Georg a brother or a cousin?' and this makes cousin the stronger reading rather than 'a different Leibholz family'. WHAT TO DO: the correction's wording should be revisited — not reversed, because the dates still rule out brotherhood, but its stated REASON is now half wrong. And the 1874 marriage record, if Bydgoszcz finds it, would name Sally's parents and make the cousin test answerable. FIXED THE SAME DAY — AND THE ROW OVERSTATED ITS OWN NOVELTY, WHICH HAS TO BE SAID. The Leibholz page's NEXT PANEL, 'A likelier reading', has carried the right answer since the start: 'Sally Leibholz was himself born at Hammerstein, on 7 JULY 1847, and married at Szubin on 24 NOVEMBER 1874. So the family had people in both towns... Georg belonged to the Hammerstein branch Sally came from — a nephew or a cousin of Jacob's, not a brother', with his status given as 'right surname, RIGHT FAMILY, wrong generation'. And open question 5 is already headed 'Is Georg a brother or a cousin?'. SO THE ARCHIVE HAD THE READING. What was wrong was narrower and sharper than this row claimed: TWO PLACES CONTRADICTED IT. (1) corrections.json entry 5 said 'disputed — A DIFFERENT LEIBHOLZ FAMILY' and gave as its reason 'two hundred kilometres apart'. (2) The Leibholz page, one paragraph ABOVE the panel that says the opposite, said 'These are two Leibholz families. Ours is the Schubin one.' BOTH ARE NOW REWRITTEN. The correction reads 'NOT A BROTHER — the dates forbid it. Probably a cousin or nephew of the same family', and rests on SIXTY-TWO DAYS rather than on the distance; the page says in as many words that it used to claim two families and that this was wrong. AND THE FIX PAID FOR ITSELF: the panel also gave the MARRIAGE DATE, 24 NOVEMBER 1874, which is SEVEN WEEKS AFTER Prussian civil registration began on 1 October. The Bydgoszcz letter drafted an hour earlier raised that boundary as an open question. It no longer speculates — it gives the date, says the record should be an ordinary 1874 Standesamt Szubin entry, and keeps the Jewish community register only as a fallback. FOURTH TIME TODAY. The cause is the same one named this morning: I read the work list and the top of a page, and the answer was further down it.

Oskar is line 54 of page 161 of the “4th Large Transport of the Elderly” (I/90), 17 March 1943, reference 15510042, FIFTY-ONE DOCUMENTS and 1,200 listed people. HIS WIFE GERTRUD SARA LEIBHOLZ NÉE LEIBKE, born Berlin 13 July 1894, IS NOT ON THAT PAGE, and Oskar's line carries no “Ehepaar” bracket although the same sheet uses one for other couples. That is suggestive and it is not an answer — she could be on any of the other fifty sheets. She has FOUR Arolsen person records of her own, mirroring Oskar's four, so the index knows where she is; the work is to open them and see which series they sit in. This matters because the site currently gives her death as 27 February 1943 in Berlin, which is BEFORE the 17 March transport — so if she is on the list, something has to give. ANSWERED THE SAME DAY IT WAS WRITTEN, AND THE ANSWER IS NO. DocID 12663551, in 1.2.4 Various Organizations, is a KULTUSVEREINIGUNG BERLIN ZÄHLKARTE for Gertrud: "Leibholz geb. Leibke · Gertrud · 13.7.94, Berlin · verh." and, in the column headed ZIELLAND — country of destination — "VERSTORBEN 27/2.43". Household address MARBURGERSTR. 13, written Berlin 8 March 1943, stamped Kataster-Verwaltung, serial 015881, with an ITS stamp "CARDED ON 6 OCT 1947" and a pencilled cross-reference to T/D 52.972 on the back. SHE DIED EIGHTEEN DAYS BEFORE THE TRANSPORT LEFT, so she cannot be on any page of it, and Oskar's line carries no "Ehepaar" bracket because by then there was no pair. IT ALSO DOCUMENTS A DATE THE SITE HAD ONLY FROM THE GEDENKBUCH — a compilation — and puts it on a contemporary card written nine days later by the office that knew her. THE FORM IS AN EMIGRATION CARD WITH "EMIGRATING" CROSSED OUT AND A DEATH WRITTEN INTO THE BOX FOR THE COUNTRY SHE WAS GOING TO. No cause of death is on it and none is inferred here, although 27 February 1943 in Berlin was the day of the Fabrikaktion. The other fifty sheets of the transport list stay unread and there is no longer a reason to read them for her.

DocID 11214358, image 001. The right-hand “Wohnungen” column carries a dated run of Berlin addresses from the 1910s onward — the household's whole moving history on one card, and the thing the Berlin directories can only sample at one household per year. IT HAS NOT BEEN READ. The scan is a grey microfilm frame 976 pixels wide and the hand is Kurrent; enlargement recovers the dates and not the street names. Two routes: ask the Arolsen Archives whether a better scan exists, or put the crop in front of somebody who reads Kurrent. Recorded as unfinished rather than as read — see data/leibholz-registration-card.tsv. TRIED PROPERLY ON 20 SEPTEMBER AND IT IS BLOCKED ON THE IMAGE, NOT ON THE EFFORT. Two processing passes were run on the crop: median-filtered denoise with five-times LANCZOS upscaling and light contrast, and a seven-times pass with an unsharp mask. THE SHARPENING AMPLIFIES THE FILM GRAIN INTO THE HAND and destroys it; the gentle pass leaves words that can be guessed at and not read. AND THERE IS ONLY ONE SCAN TO WORK FROM. Six URL variants were tested against collections-server.arolsen-archives.org — the H, V and L partitions, a _big suffix and a .tif — and every alternative returns 404. The 976-pixel frame is all the server has. WHAT WAS RECOVERED, AND IT IS ONLY THIS: the column carries a dated run BEGINNING IN 1913 — entries reading 18.13, 19.13, 21.13 and 23/4.15 — and the latest annotation on the card is Auguste's 18.1.44. THE STREET NAMES ARE NOT READABLE AND NONE IS PUBLISHED. TWO ROUTES REMAIN AND NEITHER IS THIS ARCHIVE'S TO TAKE ALONE. (1) ASK THE AROLSEN ARCHIVES whether a higher-resolution scan of DocID 11214358 exists — they hold the film, the web copy is a derivative, and they answer written requests. (2) PUT THE CROP IN FRONT OF SOMEBODY WHO READS KURRENT. Re-owned to David-or-Arolsen and marked blocked rather than next, because nothing further can be done from this end.

Found in the 43-topic Leibholz sweep: “Tracing and documentation case no. 161.094 for MEYER, GERTRUD born 09.03.1899”, indexed under the MAIDEN NAME LEIBHOLZ. The Schubin siblings this archive has are Samuel 1877, Georg 1884, Jacob 1884 and Oskar 1887. A Leibholz daughter born 1899 would fit the same family and there is no sister in the tree at all. THE CASE IS IN THE RESTRICTED T/D REPOSITORY, so the lead is an index entry and a married surname. The reachable test is her person records and the Berlin directories under MEYER — and the Schubin civil registers, which is the same Bydgoszcz thread as request 30. TESTED AND KILLED THE SAME DAY. Her second person record gives a PLACE OF BIRTH and it is GLOWITZ — Główczyce in Pomerania, KREIS STOLP. STOLP IS WHERE WILLI LEIBHOLZ WAS BORN, on the 1946 Jewish Agency questionnaire read earlier the same day. She is a Pomeranian Leibholz, not a Schubin one, and she is not Jacob's sister. THE NEGATIVE IS WORTH MORE THAN THE HYPOTHESIS WAS: the two Leibholz families are now separable by place — Schubin in Posen against Stolp in Pomerania — and that is the only thing that will ever separate them. Her fate is on the record too: DocID 127212709, "Wave 48 — 34th OSTTRANSPORT TO CC AUSCHWITZ, 4 MARCH 1943", thirteen days before Oskar's Alterstransport to Theresienstadt and a different kind of list altogether. Struck rather than done, because the row asked a question whose premise turned out to be wrong.

T/D cases 90.465 to 90.474 are one enquiry about one household: ALFRED 04.12.1887, ELSE 20.07.1889, MARIANNE 26.04.1921, LISELOTTE and KURT match Mapping the Lives at Levetzowstraße 13 to the day. HANS (90.465), GERHARD (90.466) and WALTER (90.473) sit in the same run and are NOT in the census record for that flat. Older children already out of the house, or relatives named in the same enquiry — not asserted either way. Also here: T/D 90.474 IS NOT DERRICK'S FATHER. It is Alfred's son Kurt, born 04.11.1927, and the note is written down so a later reader does not repeat the mistake. NOT ON THE CRITICAL PATH — this is the Levetzowstraße family, not Schubin — but it is the largest single Leibholz household in the archive and it is one search away. DONE 20 SEPTEMBER, BY SEARCHING THE 1939 CENSUS DATABASE ON STREET AND HOUSE NUMBER INSTEAD OF ON NAME — AND ONE OF THE THREE WAS THIS ARCHIVE'S OWN MISS. LEVETZOWSTRASSE 13 HOLDS SIX LEIBHOLZ, NOT FIVE: Alfred 04.12.1887 † 08.08.1942 · Else née Seelig 20.07.1889 † 17-19.08.1943 · Marianne 26.04.1921 · WALTER 20.09.1922 † before 08.05.1945 · Lieselotte 16.11.1923 † 06.12.1943 · Kurt 04.11.1927 † before 08.05.1945. T/D 90.473 WALTER is the fourth child and data/mappingthelives-leibholz.tsv simply did not list him; that file now carries the correction. HANS (90.465) AND GERHARD (90.466) ARE NOT AT THAT ADDRESS. The only Gerhard in the database is Dr Uri Gerhard Leibholz, born 15.11.1901, who DIED AT GÖTTINGEN IN 1982 — a survivor, and on the public record the constitutional lawyer. AND THE ONLY HANS TURNED OUT TO BE WORTH THE WHOLE ROW. HANS LEIBHOLZ, born 25.01.1912 IN PREUSSISCH FRIEDLAND / SCHLOCHAU / POMMERN, "Race" JJJJ, at DANCKELMANNSTRASSE 44 CHARLOTTENBURG on census night, DEPORTED 12 JANUARY 1943 AND DEAD AT AUSCHWITZ ON 22 FEBRUARY 1943. That is the "Hans Israel Leibholz born 1912 at Preussisch Friedland" the Yad Vashem sweep found and could not place. His flat holds three more: his mother ELSE née ROTHENBERG b.23.10.1884 IN SCHLOCHAU (no death recorded), his sister LISELOTT b.22.10.1920 in Preussisch Friedland † Auschwitz, and LAURA née ROTHENBERG b.23.09.1884 — a second Rothenberg woman of 1884 at the same address. The T/D run 8.891/8.892/8.893 is this flat, with two crossovers recorded rather than tidied. AND IT IS THE FIRST BERLIN LEIBHOLZ HOUSEHOLD FROM OUR OWN KREIS. Schlochau and Preussisch Friedland are KREIS SCHLOCHAU, and so is HAMMERSTEIN, where SALLY LEIBHOLZ was born on 7 July 1847. Every other household placed so far is Schubin, Stolp or Berlin. A KREIS IS NOT A FAMILY and this is not claimed as kinship — what it does is say where to look, and the answer is the Hammerstein and Preussisch Friedland civil registers, which are the same Bydgoszcz-region books as the Szubin letter. IT ALSO NAMES A STOLPERSTEIN: the Danckelmannstraße 44 stone, listed here for a week as belonging to "other Leibholz families", commemorates this household.

FOUND 20 SEPTEMBER: the first Leibholz survivor this archive has a name for. She filled in three Relico questionnaires in SHANGHAI in 1946 — for her brother SALLY BINDER, his wife MATHILDE, and her brother-in-law HEINRICH LEIBHOLZ. Sister to a Binder and sister-in-law to a Leibholz makes her SOPHIE NÉE BINDER, married to a brother of Heinrich. HER HUSBAND'S FORENAME IS ON NONE OF THE THREE FORMS. THREE THINGS TO DO. (1) The Berlin directory: Heinrich Leibholz is at HORNSTRASSE 19 on her form and at Neukölln, Leinestrasse 32 in the 1938-42 volumes — two addresses for one man, and the volumes will name his brothers as householders. (2) The RELICO COLLECTION ITSELF, "Questionnaires used for the search for relatives by Relico, the World Jewish Congress Relief Committee in Geneva, completed by survivors in Shanghai, 1946" — a source this archive has never used, and one that by definition records people who were ALIVE. Search it for LUWINSKI as well as LEIBHOLZ. (3) The Shanghai ghetto emigrant lists, which are held at Arolsen. THE BOUND, STATED FIRST: NOTHING TIES SOPHIE OR HEINRICH TO KURT'S FAMILY. Heinrich was born in Berlin in 1890, not at Schubin. This is a lead about a Berlin Leibholz family, and it is the first one with a living person in it. ANSWERED THE SAME DAY, AND THE HUSBAND HAS A NAME. Mapping the Lives has an EMIGRATION PLACE field. Surname LEIBHOLZ with emigration place SHANGHAI returns two people: Charlotte Abrahamsohn née Leibholz b.18.09.1889, and ALBERT LEIBHOLZ, BORN 17.05.1886 — who is the "Albert, Antiquitäten, Halensee" of the 1938 Berliner Adressbuch. Arolsen then gives him twice over. (1) DocID 131698324, a CENTRAL LOCATION INDEX card, index no. L 337909: "LEIBHOLZ · ALBERT · died 8.18.44 · 5.17.86 · Berlin · German · J · dealer · last known address SHANGHAI, 12.14.45", and under OTHER MEMBERS OF THE FAMILY ABROAD: "SOPHIE · REL. TO NO. 1: W · 6.19.87 · BERLIN". W IS WIFE. SOPHIE LEIBHOLZ NÉE BINDER WAS BORN AT BERLIN ON 19 JUNE 1887 AND SHE WAS ALBERT'S WIFE. The card also gives the journey — "arr. 6.27.39 from Berlin" — and a third person looking for them: CURT STERNGAST, c/o Silman, Rambam Str. 39, Rehavia, JERUSALEM. (2) DocID 78827389, the H.I.A.S. "lists concerning Jewish nationals who died and were buried in SHANGHAI during 1939-1948", page 24: "LEIBHOLZ, ALBERT · 5/17/86 · 8/18/44 · PARALISIS PROGRESSITIS", file L-142. SO: Berlin antiques dealer, out on 27 June 1939, dead in Shanghai on 18 August 1944, his widow still there in December 1945 and filling in relief forms in 1946. AND IT SEPARATES THE TWO SOPHIES FOR GOOD — née BINDER born Berlin 19 June 1887 and alive, against née ROTHENBERG born Schlochau 9 January 1887 and murdered. STILL AN INFERENCE AND THE ONLY ONE LEFT: that Heinrich is Albert's brother. Sophie called him her brother-in-law and she married a Leibholz; nothing yet says the two men were brothers rather than cousins. In data/leibholz-shanghai.tsv. A clean negative came with it: LUWINSKI returns nothing at Yad Vashem, victim or submitter, against a control of 66 for LEIBHOLZ.

WRITTEN 20 SEPTEMBER OUT OF ROW 200, AND IT SHOULD NOT HAVE BEEN. The row proposed writing to Poland for the Hammerstein registers — Sally Leibholz's birth of 7 July 1847, and Georg's of 7 June 1884 — and said to check first which archive branch holds Człuchów. IT WAS CHECKED, AND ASKED, AND ANSWERED, SIX DAYS EARLIER. A letter went to Archiwum Państwowe w Koszalinie on 14 SEPTEMBER, thread 1a0a1a01f08ce480, asking for exactly those two records and the synagogue books. It was ANSWERED ON 16 SEPTEMBER, reference O-S.6342.406.2026, signed by Krzysztof Skrzypiec, Head of the Słupsk Branch: the holdings contain NO registers of Standesamt Hammerstein FOR 1874 TO 1899 and NO SYNAGOGUE-COMMUNITY BOOKS FOR CZARNE AT ALL, and no information about what became of either. It was acknowledged on 17 September as requests/28-SEND-THIS-to-koszalin.txt. All of that is in site/src/data/letters.json and in row 88. WHAT TODAY'S WORK ADDS IS ONE LINE OF CORROBORATION AND NOTHING ELSE. The public finding aid says the same thing independently: szukajwarchiwach holds fond 27/364/0, "Standesamt Hammerstein [Czarne]", at ARCHIWUM PAŃSTWOWE W KOSZALINIE ODDZIAŁ W SŁUPSKU, 127 units all digitised — and its own date facet reads 1900+ : 128, 1800-1899 : 0. A search for "Synagogengemeinde Hammerstein" returns nothing, against a control of 127 for "Standesamt Hammerstein". So the branch's written negative is confirmed by its own catalogue, which is worth having and is not worth a row. FamilySearch was checked too and has filmed Jüdische Gemeinde Matrikel for Stuhm, Flatow, Marienwerder, Rosenberg, Marienburg and Frauenburg — AND NOT FOR HAMMERSTEIN. THE CAUSE, BECAUSE IT IS THE SIXTH TIME TODAY. The row was written off the back of a finding without reading letters.json or the requests directory. The finding was real; the action proposed for it was already closed. THE CHEAPEST ROUTE TO THE GENERATION ABOVE SALLY IS UNCHANGED AND IS ALREADY DRAFTED: the SZUBIN MARRIAGE OF 24 NOVEMBER 1874, which names the parents of both spouses, in requests/30-SEND-THIS-to-bydgoszcz.txt, unsent.

NEW 20 SEPTEMBER, out of the transport list. Jacob Leibholz is on the 30th Osttransport of 26 February 1943 at TREUCHTLINGERSTR. 4, Berlin-Schöneberg, and the AJDC card adds "b/SPIER" — lodging with a family called Spier. THE 1939 CENSUS HAS HIM AT MERANER STRASSE 2 WITH AUGUSTE AND SIEGFRIED. Something happened in between and this archive does not know what. THREE THINGS ARE REACHABLE WITHOUT ANYBODY'S HELP. (1) MAPPING THE LIVES BY ADDRESS — the method that found the Danckelmannstraße household. Search street TREUCHTLINGERSTR, number 4, and the whole flat comes back, SPIER INCLUDED. (2) THE BERLINER ADRESSBUCH for Treuchtlingerstraße 4, 1939 to 1943, which gives the householder — and if the householder is Spier, that settles whose flat it was. (3) The same two questions for AUGUSTE: her registration card carries an annotation dated 18.1.44 giving a street and a number that has not been read, so where she went after he was taken is also on a document. WHY IT MATTERS AND WHY IT IS NOT URGENT: it does not change a date or a name in the register. It is the last stretch of Jacob's life and it is the only part of it that is still blank. PARTS 1 AND 2 DONE WITHIN THE HOUR, AND PART 1 ANSWERED MORE THAN IT ASKED. TREUCHTLINGER STRASSE 4 ON CENSUS NIGHT HELD SEVENTEEN PERSECUTED PEOPLE FROM ELEVEN FAMILY NAMES — Brück, Cohn, Eisenberg, Eisenberger, Kaphan, Kochanowska, Löwenheim, Podbielski, Rosenstock, Sundheimer and SPIER. The Spiers are named: MAX b.12.04.1864 † 26 MARCH 1941, HANS b.02.12.1894, ALICE née WOLLSTEIN b.03.11.1899, the last two dead before 8 May 1945. The "b/Spier" on Jacob's card is a household with three names in it, and by the time he was lodging there the eldest was already dead. Eleven surnames at one address in May 1939 is not an ordinary apartment block; Jacob was moved into a house that was already full. He is not in the 1939 list for it because that night he was still at Meraner Straße 2. AND THE SAME SEARCH ON MERANER STRASSE 2 SOLVED SOMETHING ELSE. Fourteen people at that address, and one of them is ANNA PUFF, BORN 1 JUNE 1895 — who is one of the eleven names indexed on DocID 11214358, the Leibholz registration card. SO THE SECOND IMAGE OF THAT DocID IS THE HOUSE LIST FOR MERANER STRASSE 2, which is exactly what the stamp on the first image says: "Lt. Hausliste 1945 n. erfaßt". One exact match out of seven unplaced names, and the other six would not be in a database of persecuted people if they were not Jewish — a strong reading, not a proof, and recorded as such. It also gives Kurt's name on that sheet a reason: an absent son written against his parents' flat. THIRD ACCOUNT THIS ARCHIVE HAS GIVEN OF THAT DocID, offered as the best one and flagged as the third. PART 3 REMAINS: Auguste's 18.1.44 annotation is in the same unreadable Kurrent as the address column, so it is blocked with row 198. || 21 SEPTEMBER, AND MOST OF IT IS ANSWERED. The 1941 volume, Teil IV printed page 1615, gives the house: owner EISENBERG, G. SARA, WW. — a widow carrying the compulsory Sara — with tenants including STEPHAN, H. ISRAEL, ROSENSTOCK, L., and SPIER, H., HANDELSVERTRETER. THAT IS THE SPIER OF THE TRANSPORT LINE: Jacob's deportation address reads "Treuchtlingerstr. 4 b/Spier", bei Spier, and here is Spier in the building two years earlier, with a trade. A Jewish-owned house filling with Jewish tenants in 1940 is what a Judenhaus was, and the directory shows it without using the word. AND JACOB IS NOT IN THE 1941 ENTRY — that volume was compiled in the autumn of 1940, so he had not been moved there yet. THE WINDOW IS NOW BOUNDED AT BOTH ENDS: Meraner Straße 2 on the census of 17 May 1939, absent from Treuchtlinger in autumn 1940, at Treuchtlinger 4 b/Spier on 26 February 1943. WHAT IS LEFT is the 1942 and 1943 volumes, and the OCR will not give up those two pages — a fuzzy street search returns only cross-references from neighbouring streets and a combined EISENBERG + SPIER search returns nothing in either volume. They must be found by eye. PUT TO THE BERLIN DIRECTORIES 22 SEPTEMBER AND NOT FOUND. “Spier” AND “Treuchtlinger” returns 3 pages and on none of them is a Spier AT that address: the Spiers are a Konfitüren-Geschäft in W 30, a Fritz Spier of Schöneberg, and a Friedrich and Liselotte Spier of Dianastraße. THE CO-OCCURRENCE IS PAGE-LEVEL, NOT ADDRESS-LEVEL — the same trap as the two Lutherstraßen. The Spier household Jacob lodged with is not reachable by full text and needs the street section read at the image. 11 OCTOBER — AND THE FIRST THING THIS ROW HAS TO SAY IS THAT HALF OF IT WAS REDISCOVERED. THE LEIBHOLZ PAGE HAS CARRIED THE 1939 TEIL IV ENTRY FOR THIS HOUSE SINCE 29 SEPTEMBER, with the owner read as GERTRUD EISENBERG NÉE HAAKE and Spier H. identified as HANS SPIER, born 2 December 1894, with MAX on the dash line under him. This row still described the question as open and nobody had read the row against the page that answered it. AND THE FIRST ATTEMPT AT THIS CORRECTION WAS ITSELF WRONG: the 1941 volume was ALSO already read, on 21 September, quoted in full on the same page with “Eisenberg, G. SARA, Ww.” and the observation that a Jewish-owned house filling with Jewish tenants is what a Judenhaus was. TWO CORRECTIONS TO ONE FINDING IN ONE HOUR. WHAT IS GENUINELY NEW IS THREE VOLUMES — 1940, 1942 AND 1943 — AND TWO OF THEM ANSWER A QUESTION THE PAGE ITSELF ASKED: the 21 September panel ends “those two pages have to be found by eye”, and they have been, and both read E UNGENANNT. 1940 is where SARA first appears, which dates the marking to the first volume that could carry it. Not through Mapping the Lives, whose API needs authentication, but through the ADRESSBUCH STRASSENTEIL, which cost nothing and gave five years instead of one snapshot. SPIER, H., HANDELSVERTRETER, WITH A TELEPHONE, AT TREUCHTLINGER STRASSE 2.4 — the same line in the 1940, 1941, 1942 AND 1943 volumes. The house is 2 and 4 together on a corner, cross-referenced to Barbarossastraße 36a and in 1939 to Luitpoldstraße 25, so the transport list's “Treuchtlingerstr. 4” and the directory's “2. 4” are one door. AND THE OWNER GOES FROM A NAME TO A NAME-PLUS-SARA TO NOBODY: 1939 “E Eisenberg, G., Ww.” · 1940 and 1941 “E EISENBERG, G. SARA, Ww.” · 1942 and 1943 “E UNGENANNT”. The middle name dates her as Jewish on the state's own record — the decree took effect 1 January 1939, which is why the 1939 volume, compiled from 1938 returns, does not carry it. WHAT “UNGENANNT” MEANS IS NOT PROVED and this archive does not pretend otherwise; what is recorded is that the house Jacob was deported from had a named Jewish owner in the 1941 volume and none in the 1942 one. One tenant leaves at the same boundary, KEPHAN H. ISRAEL, Einrichter, in 1940 and 1941 and gone from 1942. AND SPIER CARRIES NO “ISRAEL” IN ANY OF THE FOUR YEARS, while the same page marks Eisenberg and Kephan and the same volume marks four men in one house round the corner in Martin-Luther-Straße — so the absence is evidence and not silence, by the same reasoning row 257 used for Günther's “ev.” NOT CLAIMED: that Spier hid him or sheltered him. Forced relocation into designated rooms was ordinary administration in Berlin by 1942 and “b/Spier” records an address, not a kindness. AND THE 10 OCTOBER NEGATIVE IS EXPLAINED RATHER THAN OVERTURNED: the full-text sweep found Spier and Treuchtlinger co-occurring only at page level, which was right — THE STREET PAGE DOES NOT CONTAIN THE STRING “TREUCHTLINGER” AT ALL, because the Teil IV heading OCRs as “Tremhtliagrr Str.” in 1942 and “Trerlchülürger Str.” in 1943. A search for both terms could never have found it. What works is WALKING THE RUNNING HEADS of Teil IV, which survive the OCR; the pages are order 5933, 6069, 6106, 6181 and 6088. STILL OPEN: when he moved from Meraner Straße 2 and whether Auguste and Siegfried moved with him — the directory lists householders, not lodgers, and cannot answer it. In data/treuchtlingerstrasse-2-4.tsv.

OPENED AND CLOSED IN ONE SITTING, 20 September. Geneteka was searched earlier the same day, returned nothing for LEIBHOLZ, and the KOWALSKI control ALSO returned nothing — so the negative was correctly withheld and the row left hanging. THE CAUSE IS ONE CHARACTER. Geneteka's province codes are positional: 01ds dolnośląskie, 02kp KUJAWSKO-POMORSKIE, 03lb lubelskie, 04ls LUBUSKIE. The query used w=04kp, which is not a code at all, so the select fell back silently to dolnośląskie. A SECOND TRAP IN THE SAME SITE: the results table is filled by AJAX, so a plain fetch of the search URL returns the page shell with an EMPTY TABLE that reads exactly like a zero. The rows exist only in a rendered browser. WITH THE CODE FIXED THE CONTROL PASSES and the answers are: LEIBHOLZ zero in kujawsko-pomorskie and zero in wielkopolskie, births and marriages. SCHMUL eighteen marriages in wielkopolskie, all fuzzy variants, NONE AT SZUBIN. AND THE STRUCTURAL REASON, WHICH IS THE ACTUAL ANSWER: Szubin IS in Geneteka — parish 13066, from 1715 — but it is THE CATHOLIC PARISH. Geneteka marks Jewish registers "(mojż.)" and KUJAWSKO-POMORSKIE HAS NO (mojż.) PARISH AT ALL, across all 358 entries in its list. A Jewish or civil marriage at Szubin in 1874 cannot be in this index however it is searched. STRUCK RATHER THAN DONE: the route was never going to work, and the only route to the 1874 marriage remains requests/30-SEND-THIS-to-bydgoszcz.txt, drafted and unsent. In data/schmul-and-thimm.tsv.

FOUND 20 SEPTEMBER, after Geneteka was proved structurally incapable of holding Jewish Szubin records. THE 1874 CIVIL REGISTERS OF SZUBIN ARE FILMED AND THE FILM IS ONLINE. Archiwum Państwowe w Bydgoszczy holds them in three fonds — 6/1792/0 SZUBIN MIASTO 1874-1935 with marriages from 1874 AND SKOROWIDZE (alphabetical indexes) FOR 1874-1878, 6/1795/0 Wieś I and 6/1796/0 Wieś II. FamilySearch filmed them from Bydgoszcz: catalogue koha:402152, Standesamt I, film note "Geburten, Heiraten, Tote 1874-1876", DIGITAL FILM 007998644, 1,121 IMAGES ; and koha:402191, Standesamt II, 1874-1875, DIGITAL FILM 008016003, 1,160 IMAGES . BOTH OPEN IN THE VIEWER WITH NO RESTRICTION. NEITHER IS IN THE FULL-TEXT INDEX — /service/search/fulltext/search/groupNumber returns an empty id list for both — so they cannot be searched by surname and must be browsed. WHY IT IS TRACTABLE: Prussian civil registration began on 1 OCTOBER 1874, so the 1874 volumes cover THREE MONTHS. The marriage book should hold a few dozen entries and the skorowidz a single page. The target is SALLY (SALI, SALOMON) LEIBHOLZ and FRIEDERIKE SCHMUL, 24 NOVEMBER 1874. WHICH STANDESAMT IS NOT SETTLED: the town is the likelier, but Standesamt I's catalogue title says 1875-1876 while its own film note says 1874-1876 — the title and the film disagree and the film is what was photographed. BOTH must be checked. AND IT CHANGES REQUEST 30, WHICH HAS BEEN UPDATED. The letter now cites the three fond references so Bydgoszcz does not have to find them. If the browse succeeds the letter is only needed for a certified copy, if at all. In data/szubin-civil-registers.tsv. WITHDRAWN WITHIN THE HOUR, AND THE ERROR IS MINE. This row first said "BOTH OPEN IN THE VIEWER WITH NO RESTRICTION". THEY DO NOT. The claim came from the viewer reporting a page count and NOT FROM LOOKING AT THE PAGE. With the browser pane displayed, EVERY THUMBNAIL IN THE FILMSTRIP CARRIES A PADLOCK, and opening one returns FamilySearch's own notice: "IMAGE RESTRICTED. Image access is typically determined by local laws or the custodian who has the original document. To view these images do one of the following: ACCESS THE SITE AT A FAMILYSEARCH CENTER." THE LESSON IS GENERAL: AN IMAGE COUNT IS NOT AN IMAGE. FamilySearch serves a film's length to everybody, restricted or not; the padlocks exist only in the rendered page. Availability cannot be read from metadata. WHAT SURVIVES: the registers exist, they are at AP Bydgoszcz in three named fonds, the town fond has SKOROWIDZE FOR 1874-1878, and the films are 007998644 and 008016003. Only the viewing is blocked, and by the CUSTODIAN rather than by FamilySearch. AND IT GIVES DAVID A FREE ROUTE THIS ARCHIVE CANNOT TAKE. The restriction lifts inside a FAMILYSEARCH CENTRE OR AFFILIATE LIBRARY, of which there are several in Brisbane, and access is free. Film 007998644, the 1874 skorowidz, the marriage of SALLY (SALI, SALOMON) LEIBHOLZ and FRIEDERIKE SCHMUL on 24 NOVEMBER 1874 — an afternoon at a terminal. Re-owned to David and marked blocked. REQUEST 30 TO BYDGOSZCZ IS THEREFORE NECESSARY AFTER ALL, and it is a better letter for citing the three fond references. AND THE POLISH SIDE WAS CHECKED TOO, SO THE PICTURE IS COMPLETE. Film 008016003, Standesamt II, IS PADLOCKED AS WELL — the restriction is on the Szubin material, not on one reel. And SZUKAJWARCHIWACH HAS NO SCANS OF IT EITHER: the portal returns the three USC Szubin fonds AT FOND LEVEL ONLY, with no unit rows, no inventory and no scan counts, across four differently-worded searches. THE CONTROL THAT MAKES THAT A FINDING: the same search box returns 127 UNIT-LEVEL ROWS WITH "Number of scans" for "Standesamt Hammerstein" at Koszalin/Słupsk. The portal does show units and scans when they exist; it has neither for Szubin. SO: THE REGISTERS EXIST AND NOBODY CAN SEE THEM FROM A DESK. Two routes remain and both need a person — a free terminal inside a FamilySearch Centre in Brisbane, or the letter to Bydgoszcz.

FOUND 20 SEPTEMBER in the Geneteka spelling sweep David asked for. Standesamt OLSZTYN-ALLENSTEIN, birth entry no. 11 of 1874: MARTHA LUWENSKI, FATHER JOSEPH, MOTHER ELISABETH BILITEWSKI. Entry eleven of a registry that opened on 1 October 1874. JOACHIM LUWINSKI WAS BORN AT ALLENSTEIN ON 21 OCTOBER 1865. A Joseph of this surname registering a daughter in that town in 1874 is the right place, the right registry and the right generation to be HIS FATHER OR HIS ELDER BROTHER — and the family tree already holds a JOSEPH LUWINSKI with no dates and no place, inside the hundred-name Polish block this archive has always called a surname index rather than a descent. NOT PROVED AND NOT ASSERTED. What makes it worth a row is that it is the FIRST TIME that block has touched an actual record. WHAT TO DO: the Allenstein Standesamt registers are at ARCHIWUM PAŃSTWOWE W OLSZTYNIE. Birth no. 11 of 1874 would name Joseph's age, trade and residence; and if Joachim's own birth of 21 October 1865 is in the Catholic register of Allenstein before civil registration, it would name his parents outright. Also search BILITEWSKI, which is a rarer name than Luwinski and may carry the family further. AND ONE MORE GENETEKA PASS IS OWED: only the surname was swept. The forenames JOACHIM and JOSEPH, and the maiden name BILITEWSKI, have not been.

FOUND THE SAME WAY. RESZEL-RÖSSEL USC: 1876 no. 68 FRANZ LOWINSKI · 1881 no. 31 ALBERT LOWINSKI · 1884 no. 9 BERNHARD LOWINSKI. The Geneteka index does not carry their parents. KREIS RÖSSEL IS THE DISTRICT SAUERBAUM IS IN. Johanna Thimm was born there in 1871 and Auguste Luwinski in 1893. A household of this surname is registering children in that district from 1876, which is between those two dates. NOT KIN ON ANY EVIDENCE HELD. What it does is tell the Olsztyn archive exactly which three entries to pull, and those entries carry the father's name, age, trade and village — which would say at once whether this is Joachim's family or a different Lowinski. NOTE THE SPELLING: the same registry spells one woman LOWINSKI in four years and LUWINSKI in two. The forms are not different families and must not be treated as such.

NINE FOLIOS, CANBERRA, PAPER, NOT DIGITISED, access status "not yet examined". This is the same nominal roll Arolsen holds: BGO 65/135 lines 85 to 88 against the NAA control symbol 1965/86/85-88. IT ALREADY CORRECTED THIS ARCHIVE ONCE — its title says the family came by SHIP, the FLAVIA, ex Bremerhaven, arriving 17 September 1965, and this archive had published that they flew. THE FREE MOVE FIRST: an ACCESS EXAMINATION request to the National Archives of Australia costs nothing and is made by email. It does not produce a scan, but it opens the item and it is the step that has to happen before any copy can be ordered. Ask for examination of item ID 5435314, and at the same time for A714 18/8913, A435 1944/4/1672 and A1478 EPCP63.

SIX NAA FILES, none scanned. Born Berlin 20 August 1903; landed Melbourne per ANCHISES 12 June 1939 with Tilde (or Hilde) and a child; naturalised 27 July 1944 at South Yarra, Victoria, PREVIOUS NATIONALITY STATELESS. AND ONE OF HIS FILES IS ABOUT PROPERTY INSIDE POLAND: A1478 EPCP63, "German property taken over by the Polish Government - Leibholz J P Mr", 1947, with a claim following at A1458 EPCH18 and German currency bonds at A1444 EPC164 — the last of those filed under the spelling LAIBHOLZ. OUR LEIBHOLZES ARE FROM SCHUBIN IN POSEN. This is the first document ever found that puts a Leibholz, a Berliner and Polish property in one sentence. NOT KIN ON ANY EVIDENCE HELD, and kept as its own lane. The cheap test is the Victorian death index and the Melbourne Chevra Kadisha, both of which would give his parents. || DONE 21 SEPTEMBER, on the free Victorian historical index. REG. 26252/1971: "LEIBHOLZ, Joachim Peter, Death, mother NANETTE, mother's family name at birth NETTER, father LEIBHOLZ WILLIAM, place of birth BERLIN GERMANY, place of death TOORAK, age at death 68." He stayed: landed 1939, naturalised at South Yarra 1944, died thirty-two years later two suburbs away, aged 68 and consistent to the year with a birth on 20 August 1903. AND THE ARCHIVE HAS ALREADY MET A WILLIAM LEIBHOLZ OF BERLIN — the 1920 directory has "WILLIAM LEIBHOLZ, GENERALDIREKTOR and STADTRAT", and the 1913 name section has "William, Kaufm., W15 Kurfürstendamm 169", the only William on that roll. STRONG, NOT PROOF; the Victorian death certificate carries the parents in full and is the thing that would settle it. IT DOES NOT CONNECT HIM TO THIS FAMILY — ours are Konfektion merchants from Schubin and William is a managing director on the Kurfürstendamm. The Melbourne lane stays a lane. THE SWEEP WAS DONE PROPERLY: the index takes a wildcard, so L*HOLZ was run across births, deaths and marriages 1900-1995 and returns exactly this one record — with the honest limit that Victorian birth indexes close about 1922 and marriages about 1942, so the child off the Anchises could not appear whatever happened.

FOUND BY RUNNING LOWINSKI, WHICH HAD NEVER BEEN RUN. Eight documents in the South African Government Gazette: THOS. H. LOWINSKY 1922 · LOWINSKY, M. 1939 · MARK LOWINSKY 1944 and 1961 · RUPERT ESMOND LOWINSKY, retired, of Riebeek, District Maclear, deceased estate 671/55 with a first and final account advertised 30 September 1955 · 3747/84/3A in 1985. NOT OURS, AND THE REASON IS A DATE: Thos. H. Lowinsky is in print in 1922 and Kurt Luwinski did not reach South Africa until 1933. PARKED, NOT PURSUED. It is on the register of lines so that nobody rules it out a second time. If the two names ever have to be separated properly, estate 671/55 names his heirs.

Item 8 of the twenty. THE ROLLS FOR THOSE YEARS ARE BEHIND A SUBSCRIPTION. FamilySearch's free Australian electoral rolls stop at 1949 and the family did not land until 1965; the 1903-1980 set is Ancestry's, and an Ancestry subscription is a purchase this archive does not make. NOT A NEGATIVE RESULT — an unopened instrument. The free substitutes are the NAA files at row 208, the Queensland death certificate, and Australian telephone directories at the State Library of Queensland.

Item 6 of the twenty. The search application hung on "Loading...", and its own configuration endpoint then began answering "No API key found in request" to this client. NOT A RESULT EITHER WAY: nothing was searched, so nothing is ruled out. AND NOTE WHAT TROVE IS ACTUALLY FOR HERE: copyright closes Australian newspapers at 1954 and this family landed in 1965, so the newspapers cannot hold them. THE GAZETTES CAN — the Commonwealth of Australia Gazette runs on past 1954 and carried naturalisations. Retry, restricted to the Gazette category, across every spelling. || RETRIED 21 SEPTEMBER, A DAY LATER, AND IT IS THE SAME. The application loads, the search box takes text, Return and the submit button both register — and "Search results for:" stays blank and no search request is ever sent. REPRODUCIBLE ACROSS TWO DAYS AND TWO SESSIONS. THE HONEST BLOCKER, NAMED: Trove's own endpoints answer "No API key found in request" to this client, and a Trove API key is free but requires REGISTERING AN ACCOUNT, which this archive does not do on David's behalf. So this is not an instrument that is down; it is one that needs a key, and getting the key is David's to do in about two minutes. Everything else about item 6 stands: it is the Commonwealth Gazette that matters, not the newspapers, because copyright closes those at 1954 and this family landed in 1965. 22 SEPTEMBER — IT WORKED, AND IT CORRECTED THIS ROW. No fix, no workaround: the application simply rendered. The one obstacle was a CULTURAL ADVICE overlay, closed rather than permanently dismissed. AND THE ROW'S PREMISE WAS WRONG. It said Australian newspapers close at 1954 so only the gazettes could hold this family. THE CANBERRA TIMES IS DIGITISED 1926 TO 1995, and every Luwinski hit Trove returns is a Canberra Times article from after 1954. LUWINSKI returns NINE in Newspapers & Gazettes and nothing in the other categories. THE THREE PREVIEWED ARE ALL WITHHELD: they concern a Canberra household of the surname in the 1970s and 80s, one of them a court report, and at least one names a LIVING person this archive is instructed never to name. Nothing from them is published — not the names, not the address, not the substance. The remaining six were not opened, because the three that were read establish the set is about living people. The living-persons guard was run against the built site as a check on this archive: nine guarded names, none present.

A free index to DEATH, FUNERAL AND PROBATE NOTICES in Australian newspapers, running to the present, where the Queensland state index stops at 1996. MANFRED DIED AT BRISBANE ON 20 SEPTEMBER 2012 and a Courier-Mail funeral notice would NAME THE SURVIVORS. Attempted four times on 20 September 2026 and the server answered "Server Overloaded" every time. NOTHING WAS SEARCHED, SO NOTHING IS RULED OUT. Retry across all seven spellings, and separately for MARQUARDT and WAGNER, which are the surnames Helga carried. || RETRIED AGAIN 20 September, late: "Server Overloaded" for the eighth time. Not searched. || RETRIED 21 SEPTEMBER AND THE DIAGNOSIS IS SHARPER. The SEARCH PAGE NOW LOADS — the form, the fields, the notice that Ryerson is free. SUBMITTING IT RETURNS "Server Overloaded" EVERY TIME, on three attempts with pauses, and a further page load then failed the same way. So the site serves and the search backend does not, which is different from what was recorded yesterday and worth having: the block is on the query, not the host. It is an intermittent public service with no workaround available here. STILL THE ONLY FREE INSTRUMENT THAT REACHES A 2012 BRISBANE FUNERAL NOTICE. Left to the scheduled retry. 22 SEPTEMBER — FOUR MORE REFUSALS, AND THE FIRST REAL ANSWER. search.php returned “Server Overloaded” four more times, about fourteen across three days. NOTHING SEARCHED, NOTHING RULED OUT. BUT THE DIAGNOSIS IS NOW EXACT: THE REST OF THE SITE SERVES PERFECTLY — home page, coverage tables and per-newspaper detail pages all load. It is search.php ALONE that refuses, so this is a throttled endpoint and not a dead archive. AND THE COVERAGE QUESTION WAS SETTLED WITHOUT SEARCHING: BRISBANE COURIER MAIL is indexed from 28 August 1933 to the present, six days a week, 377,328 records, last updated 19 September 2026 — with DEATH NOTICES from 22 February 1943 (92,248) and FUNERAL NOTICES over the same span (247,302). Manfred died at Brisbane on 20 September 2012, in the middle of both. The instrument covers the newspaper, the year AND the notice type that names survivors. The only thing in the way is a search box that will not answer. 8 OCTOBER — IT SERVED, AND THE ANSWER IS A REAL ZERO WITH A CONTROL BEHIND IT. After roughly fifteen refusals across three weeks, search.php answered. THE INDEX STATES ITS OWN SIZE: 10,195,938 NOTICES FROM 531 AUSTRALIAN NEWSPAPERS, publisher and funeral-director sites. FOUR EXACT SURNAME SEARCHES, no year, paper or state limit: LUWINSKI 0 · LUWINSKY 0 · LOWINSKI 0 · LOWINSKY 0. MANFRED HAS NO DEATH, FUNERAL OR PROBATE NOTICE IN IT. AND THE CONTROL CAME FREE: LEWINSKI RETURNS THIRTY-TWO — Hobart, Adelaide, Melbourne, Sydney, none in Queensland, no Manfred and no Helga — so the harness finds a Polish-German surname one vowel away and the zeroes are real. WHAT IT MEANS: this row already established that the BRISBANE COURIER-MAIL is indexed 1933 to the present, 377,328 records, with death notices from 1943. THE PAPER THAT WOULD CARRY HIM IS COVERED AND HE IS NOT IN IT — so either no notice was placed, which for a man of eighty in 2012 is ordinary, or it used a spelling this archive does not hold. IT DOES NOT TOUCH THE DEATH ITSELF, which comes from the family tree. ONE OPTION TRIED AND REFUSED AS AN INSTRUMENT: “Sounds like” on LUWINSKI returns 1,010 notices under thirty-four surnames, 883 of them LAING, and NOT ONE Luwinski, Lowinski or Lewinski — it does not even return the spelling typed into it. No negative is taken from it. STILL TO DO, AND SHORT: LUVINSKI, LUBINSKI, and MARQUARDT and WAGNER with the given name HELGA; the endpoint serves about seven queries and then throttles again. The form's field names are recorded in data/australia-indexes.tsv so nobody has to find them twice.

A Queensland death certificate NAMES THE PARENTS, and Manfred's parents are what the family tree gets worst: it makes him the son of Alfred Luwinski of Przasnysz, who died at Sachsenhausen in 1942 and was a Polish Catholic farm labourer. BUT THE FREE HISTORICAL INDEX STOPS AT 1996 AND HE DIED IN 2012, so this is not an index search — it is an application for a recent record, and Queensland issues those only to people with a recognised entitlement. BEFORE ANYTHING ELSE: is there one? Manfred is not proven kin to Kurt, so a family relationship may not exist to claim. The cheap alternative that needs no entitlement at all is the Ryerson Index at row 213 and the funeral notice it points to.

Item 10 of the twenty. The 1965 emigration roll forwards the family c/o Mr H. MARQUARDT at Hackett, ACT, and Manfred's wife was HELGA ERNA MARQUARDT: the sponsor is almost certainly her family, already out. A keyword sweep of the NAA for MARQUARDT, 1950 to 1966, returns 83 items and THIRTY-ONE WITH AN H FORENAME. THE BEST FIT IS HARRY WALTER MARQUARDT, born 22 MARCH 1934, German, out on the ANNA SALEN in March 1955 — A2478 MARQUARDT HW and A2571 MARQUARDT HARRY. AND THE SAME LINE RULES OUT THE OBVIOUS READING: Helga Erna was born 1 MAY 1934, forty days later. They cannot have the same mother, so brother and sister is precisely what this pair of dates forbids. Cousins or no relation are both still open. WHAT DECIDES IT IS AN ADDRESS — the A2478 selection document carries the sponsor's, and the electoral rolls would too if they were not behind a subscription. ALSO RECORDED: two other Helga Marquardts already in Australia before ours, one naturalising in South Australia in 1960 and one Lithuanian off the General Stuart Heintzelman in 1949. Neither is ours and both are the kind of near-match that swallows an afternoon. || 21 SEPTEMBER: RECLASSIFIED AS BLOCKED, BECAUSE IT IS NOT MINE TO FINISH. The shortlist is built and cannot be narrowed without an ADDRESS, and there are exactly two places an address lives: the A2478 selection document at the National Archives of Australia, which is in the access-examination letter and therefore David's to send; and the Australian electoral rolls for 1965-80, which are behind a subscription this archive does not buy. NOTHING FURTHER CAN BE DONE HERE. Leaving it as "next" for this archive would misdescribe who is holding it up. || 27 SEPTEMBER 2026, FROM TROVE — THE HACKETT HOUSEHOLD HAS A STREET. Commonwealth of Australia Gazette, 9 March 1967, Certificates of Naturalisation: "MARQUARDT, ANNE MARIE, 2 KENNY PLACE, HACKETT, 28.9.66". This row says WHAT DECIDES IT IS AN ADDRESS, and there is now a Marquardt household at Hackett with one. IT IS NOT YET THE SPONSOR'S ADDRESS PROVED — the 1965 roll says "c/o Mr H. Marquardt, Hackett" and this is an Anne Marie, so the link is the suburb and the surname, not a document. H. MARQUARDT IS NOT IN THE GAZETTE LISTS, CHECKED ACROSS FIVE MORE: every other Marquardt is North Ryde, West Hoxton, Longreach or Montrose, none in the ACT. If the Anna Salen candidate arrived March 1955 he was eligible to naturalise about 1960, before the surviving lists, SO THE ABSENCE IS EXPLAINED AND IS NOT EVIDENCE AGAINST HIM. NEXT: 2 Kenny Place is a specific address to put against the A2478 selection document when the NAA access examination comes back. See data/trove-2026-09-27.tsv.

Items 1, 2 and 3 of the twenty, and the plan had the addressee wrong: it put these files at the Landesarchiv in B Rep. 025. UNTIL THIS YEAR THEY WERE NOT AT AN ARCHIVE AT ALL — they belonged to the ENTSCHÄDIGUNGSBEHÖRDE, a department of LABO, a live authority. ON 1 MAY 2026 THE LANDESARCHIV BERLIN TOOK THE WHOLE OF IT: about 280,000 proceedings, FIFTEEN KILOMETRES OF PAPER, stem files and subsidiary files together. The question is four and a half months old. The archive's stated minimum to ask is the name and the LIFE DATES, and this archive has those for Kurt (11.05.1912 - 15.12.1968), Auguste (28.08.1893 - 22.06.1975) and Siegfried (05.08.1925). About four weeks from enquiry to reading, with a Schutzfristverkürzungsantrag likely. AND RELATIVES MAY HAVE TEN FILES A YEAR DIGITISED FREE — the letter states the relationship exactly as the records state it, that Derrick is Kurt's son on every document, and leaves the status to them, because this archive has published its own doubt about that paternity and will not claim round it. Drafted as a SUPPLEMENT to the 14 September thread, not a new letter.

AN INSTRUMENT THIS ARCHIVE HAD NEVER HEARD OF, named on the Landesarchiv's own page: a NATIONAL card index of Bundesentschädigungsgesetz claims, held by the Bezirksregierung Düsseldorf at bzk@brd.nrw.de, with a Merkblatt at brd.nrw.de/document/20230201_1_15_BZK_Merkblatt_deutsch.pdf. WHY IT COMES BEFORE BERLIN: every western Bundesland ran its own compensation authority and § 185 BEG decided which was competent. Kurt left Berlin in 1933 and died in Johannesburg in 1968; Auguste died in Baden-Württemberg in 1975; Siegfried lived in Potsdam, on the other side of the wall. THREE PEOPLE, THREE POSSIBLE JURISDICTIONS, and a Berlin-only enquiry can return nothing while a file sits in Stuttgart. The BZK answers whether a claim was ever filed at all and where. One letter covers Kurt, Auguste, Siegfried and the heirs of Jacob.

Item 3 of the twenty, and it rides on row 216 rather than needing its own letter yet. "JAQUES LEIBHOLZ & CO, KONFEKTION, W8 KRONENSTRASSE 59.60" is in the 1938 Berliner Adressbuch and gone from every volume after it. A liquidated Jewish firm generates a property file at the WIEDERGUTMACHUNGSÄMTER VON BERLIN naming the owners, the value, the purchaser who took it and THE POST-WAR CLAIMANTS — and the claimants are the surviving family, which is the single list this archive most wants. Asked as the second half of the Landesarchiv supplement. If they answer that it is a separate holding with its own procedure, it becomes its own letter.

Item 17 of the twenty, and the reasoning is the Bydgoszcz reasoning. A MARRIAGE RECORD NAMES THE PARENTS OF BOTH PARTIES. Johanna's are known — Anton Thimm b. 16 July 1840 and Elisabeth b. 15 March 1845. JOACHIM'S ARE KNOWN TO NOBODY, and they are the oldest generation reachable on this side of the family. He was born at Allenstein 21 October 1865, she at Sauerbaum 25 March 1871, and their daughter Auguste at Sauerbaum on 28 August 1893 — so the marriage falls between about 1888 and 1893, at USC Biesowo or USC Reszel. The letter asks for that first and for the four indexed entries of rows 206 and 207 second. IT ALSO GETS THE PRE-1874 PROBLEM RIGHT: Prussian civil registration began on 1 October 1874, so the births of Joachim in 1865 and Johanna in 1871 are CHURCH records — St Jakobi Allenstein and the Catholic parish of Gross Bössau, whose register runs 1690 to 1897 and names Sauerbaum expressly — and the letter asks whether those books are at Olsztyn or at the Archiwum Archidiecezji Warmińskiej. Every spelling is listed for them: LUWINSKI, LUWINSKY, LOWINSKI, LOWINSKY, LUWENSKI, LUWWINSKY.

DONE, 20 September 2026. LUWINSKI · KURT MYSZISLAUS · 963 — the May block of the L section, page 75 of P Rep. 523 Nr. 1395, the name index to the birth register of STANDESAMT BERLIN VIII for 1912. Twenty-one candidate volumes became one: Schöneberg was excluded by reading its whole 1912 L section (entries 14 to 1119, no Luwinski), the 1913 directory gave Chodowieckistraße 32, and the Berliner Adressbuch's own street index — whose first column is headed STANDESAMT — gives houses 23 to 32 as office 8. FOUR THINGS AGREE: the surname, the forename, the month, and a district arrived at independently. AND THE ARITHMETIC CHECKS — entry 963 of a register that passes 2561 by December is 37.6 per cent through it, and 11 May is 36 per cent through the year. Every L page of 1912 was read: exactly one Luwinski in the whole year. A second forename nobody has ever seen: MYSZISLAUS, a German clerk's spelling of Mieczysław.

DONE, 20 September 2026. "LEIBHOLZ · mos. · verheiratet mit LUWINSKI · kath. · Nr. 6" and the same entry again under her name, on page 221 of P Rep. 161 Nr. 271, the marriage index of STANDESAMT SCHÖNEBERG II. The family tree gives the wedding as 5 January 1924; a register restarts at 1 each January and the entries above are 1, 2 and 4, SO THE SIXTH MARRIAGE OF THE YEAR IS A MARRIAGE IN THE FIRST WEEK OF IT — date and number agree without either being derived from the other. AND THE CLERK WROTE THE RELIGIONS: Jacob mos., Auguste kath. Her Catholicism had been inferred from the Ermland, from Sauerbaum and from a census card; here a registrar writes it at her own wedding, in the same line as her husband's Judaism. FOUND BY ELIMINATION: Berlin VIII's January 1924 L block (P Rep. 523 Nr. 1054 p.34) and Schöneberg I's opening 1924 L block (P Rep. 160 Nr. 1036 p.210) were read first and hold neither name. AND IT DATES THE MOVE: they were in eastern Schöneberg by January 1924, nineteen months earlier than any document previously showed.

A finding that was about to be published and was false. Three queries for Schöneberg returned nothing and a sweep of all 142 offices put Schöneberg, Neukölln, Köpenick and Weißensee in the empty column. EVERY OFFICE NAME WITH AN UMLAUT RETURNED ZERO AND EVERY NAME WITHOUT ONE RETURNED ROWS. The site's database stores the names DOUBLE-ENCODED — the stored string is literally "Standesamt Köpenick" — and the match is on the whole string, so the CORRECT spelling finds nothing and the MANGLED one finds seventeen volumes. Schöneberg has more online than almost anywhere: births 1874-1913 and deaths year by year to 1994. WHAT CAUGHT IT was not cleverness but noticing that all four failures shared a property and none of the successes did. Kept as a standing check: when a set of negatives has a common character, suspect the character before believing the negatives.

DONE. Both volumes found by mapping each year's Teil III pagination onto its image numbers rather than by searching for the name, which is what the row said would be needed. 1912: NUMBER 32 IS A NEUBAU — "32 Neubau. E. Berl. Terrain-Gesellschaft", owned by the developer, no tenants at all. AND NO LEIBHOLZ AT 31 EITHER: the 1912 run goes Köhler, Koettlitz, KRINKE, Liebenhagen, where 1913 reads Köhler, Koeltlitz, LEIBHOLZ, Liebenhagen. 1914: THE LUWINSKIS ARE GONE AND GEORG LEIBHOLZ IS STILL THERE — number 32's alphabetical list runs Landow, Michaelis with nothing between, and 31 still carries "Leibholz, G., Kfm." SO THE LUWINSKI TENANCY IS ONE VOLUME WIDE: into a brand-new building during 1912, out again before the autumn of 1913. AND IT QUALIFIES THIS ARCHIVE'S OWN CLAIM — the birth-entry file counted the district as one of four agreeing facts, but the address dates from the autumn of 1912 and the birth was in May, and the family had moved that year. Three agree outright, one partly. The entry itself is untouched: the register index was read.

The thing that now stands between this archive and Kurt's birth entry number. Row 220 enumerated twenty-one candidate 1912 birth-index volumes because the district was unknown. IT IS NO LONGER UNKNOWN — a Luwinski household was at Chodowieckistraße 32, PRENZLAUER BERG, POSTAL N 55, in the year he was born. Which of the old Roman-numeral Berlin registry offices covered that street in 1912 is NOT established and is deliberately not guessed at. The Berliner Adressbuch's own front matter lists the Standesämter with their districts; so does the Landesarchiv. One of those settles it, and settling it cuts twenty-one volumes to one or two. || ANSWERED THE SAME DAY: STANDESAMT 8. The Berliner Adressbuch's second part carries an alphabetical street index whose first column is headed "Standesamt". Chodowieckistraße: houses 1-10 → 8, 11-20 → 8, 21-22 → 10c, 23-32 → 8. Number 32 is office 8, and so is number 31 where the Leibholz lived. THE INSTRUMENT IS GENERAL: that table gives the registry office for every street in Berlin, and this archive will need it again.

DONE the same day, from the name section of the same volume: "— GEORG, KAUFM., NO 55, CHODOWIECKISTR. 31 II." A merchant, second floor, next door to the Luwinskis at 32. This archive already holds a GEORG LEIBHOLZ, born at Hammerstein, who cannot be Jacob's brother and has his own lane; WHETHER HE IS THIS MAN IS NOT ESTABLISHED, though the 1913 volume lists exactly one Georg Leibholz in Berlin. AND FOUR LINES BELOW HIM: "— JAQUES, KAUFM., W 62, LUTHERSTR. 34 II." The man Auguste married eleven years later, in the same column of the same page — and TWENTY-FIVE YEARS EARLIER than the archive's previous earliest address for him. The 1913 Berlin Leibholz roll is transcribed in full in the evidence file. AND AN EIGHTH SPELLING, printed by the directory itself: "LEIPHOLZ, s. a. LEIBHOLZ", never searched anywhere.

THE ONE TO SEND FIRST OF EVERYTHING DRAFTED TODAY. A single document by its exact register number: no searching, no name to match, no protection period to argue — births are open after 110 years and 1912 was 114 years ago. WHAT IT WILL CONTAIN: Auguste Luwinski's age, occupation and ADDRESS ON 11 MAY 1912, which this archive has never had; THE FATHER'S COLUMN, whether filled or empty, and either answer is worth having; the informant; and SAMTLICHE RANDVERMERKE — the marginal notes. Auguste married Jacob Leibholz on 5 January 1924, and if Kurt was legitimised then, the note is in that margin. It is a copy order and will cost something; the letter asks for an estimate first and does not request a certified copy.

The 1913 directory's own cross-reference reads "LEIPHOLZ, s. a. LEIBHOLZ" — the compiler treating the two as one name. THIS ARCHIVE HAS NEVER RUN IT: not at Arolsen, not at the Gedenkbuch, not at Mapping the Lives, not at the NAA, not at gazettes.africa, not in the Berlin indexes. Every bounded negative recorded for LEIBHOLZ · LAIBHOLZ · LEIBHOLTZ · LEIBBHOLZ is one spelling short until this is run. The 1925 volume repeats the same cross-reference, so it is not a one-year quirk. || STARTED THE SAME DAY. It is a real Berlin surname, present in every Adressbuch volume sampled from 1913 to 1943, and THE CROSS-REFERENCE RUNS BOTH WAYS — 1913 and 1924 print "Leipholz, s. a. Leibholz", 1940 prints "Leibholz s. o. Leipholz". AND ONE MAN MAY BE PRINTED UNDER BOTH: ADOLF, police officer, NO18 Friedenstraße — Leipholz in 1924 at no. 30, Leibholz in 1940 at no. 47, in the eleven-householder run this archive has already published. TEXT LAYER ONLY, NOT VERIFIED AT THE IMAGE: three crops of the 1924 page returned a different column because that page's ALTO coordinates do not map onto the served image. NAA: nothing. Gazettes.Africa: nothing. STILL TO RUN — Arolsen, the Gedenkbuch, Mapping the Lives, Yad Vashem, Queensland, Geneteka, and the Landesarchiv name indexes. Seven recorded negatives stay provisional until they are. || MAPPING THE LIVES RUN, AND THE RULE HAS PRODUCED A PERSON: WEILER, ELISABETH NÉE LEIPHOLZ, BORN 1 JANUARY 1891, BERLIN — exactly one, with LEIBHOLZ as the control on the same screen returning the hundred-capped list the archive already holds. She is in none of this archive's files under any spelling. NOT KIN ON ANY EVIDENCE HELD. Six instruments left: Arolsen, the Gedenkbuch, Yad Vashem, Queensland, Geneteka, and the Landesarchiv name indexes. || GEDENKBUCH RUN: LEIPHOLZ returns NOTHING, with LEIBHOLZ returning fifty as the control. Five instruments left. || YAD VASHEM RUNS IT AND THE RULE IS PROVED ON NAMED PEOPLE: WILLI LEIPHOLZ b.1887 Berlin and JENNY LEIPHOLZ b.1893 Berlin, both murdered, both with a Page of Testimony — AND THE GEDENKBUCH'S FIFTY CONTAIN "Leibholz, Wilhelm Willi (* 30.12.1887 in Berlin)" and "Leibholz, Jenny (* 11.02.1893 in Loosen)". The same couple, spelt with a B by a German federal archive and with a P by a Jerusalem memorial. That supersedes the unverified Adolf-the-policeman reading. Four instruments left: Arolsen, Queensland, Geneteka, the Landesarchiv name indexes. || QUEENSLAND: nothing. GENETEKA: the empty-parameter trap caught the archive a second time — both spellings first returned an empty table with province and record type left blank, and the Kowalski control with them filled in returned rows at once. RE-RUN PROPERLY, LEIPHOLZ FINDS WHAT LEIBHOLZ CANNOT IN OUR OWN PROVINCE: kujawsko-pomorskie births give Leibholz "Brak danych" and Leipholz one hit — Gottlieba Leipholz, mother, Runowo Krajeńskie 1905. AND A WARNING TO GO WITH THE RULE: warmińsko-mazurskie returns an 18th-century PROTESTANT Leipholz family at Orzysz-Arys in Masuria, which is not a Jewish family from Posen by any reading. Leipholz is also a surname in its own right. THE RULE WIDENS THE SEARCH; IT DOES NOT WIDEN THE FAMILY. Two instruments left: Arolsen and the Landesarchiv name indexes. || AROLSEN COMPLETES IT, WITH A THIRD PROOF: LEIPHOLZ returns BODIN, ELSE, maiden name LEIPHOLZ, born HAMMERSTEIN 12.08.1884, and LEIPHOLZ, HEINZ, plus three T/D cases. The Gedenkbuch's fifty contain "Bodin, Else (* 12.08.1884 in Hammerstein)" under LEIBHOLZ. A THIRD INSTITUTION, A THIRD PERSON, THE SAME TWO SPELLINGS — and this one is Hammerstein-born, which is where Sally Leibholz was born and where five of the fifty were born. NINE INSTRUMENTS SWEPT IN ONE EVENING. The rule produced FIVE PEOPLE the archive did not have — Elisabeth Weiler, Willi, Jenny, Heinz and Gottlieba — a Masurian Protestant family that must be kept out, and three separate demonstrations of the equivalence. DONE.

ANSWERED THE SAME DAY IT WAS RAISED, AND THE ANSWER IS NO. Berlin's MARTIN-LUTHER-STRASSE in Schöneberg has carried that name SINCE 28 DECEMBER 1899; its only earlier names were Heinrich-Kiepert-Straße (1893-1899) and Rostocker Straße (1884-1893). It was never a Lutherstraße. The only Lutherstraße in Berlin today is in STEGLITZ, formerly Grabenstraße. Source: berlin.kauperts.de. AND THE REFUTATION IS WORTH MORE THAN THE GUESS WAS: Jaques Leibholz's 1913 address is postal district W 62, which is BERLIN PROPER, and Schöneberg was a separate city until 1920 with its own postal districts. HE WAS NOT LIVING IN SCHÖNEBERG IN 1913 AT ALL. The family's arrival there is a move, not a continuity. Raised at 21:0x, struck within the hour, and the speculation never reached a narrative page in an unqualified form.

Sits in Gmail drafts as a reply on thread 1a0a19a926964a01, one click from going. It supplies what the 14 September letter could not: the exact register numbers. STANDESAMT BERLIN VIII, GEBURTSREGISTER 1912, NR. 963 for that letter's point 3, and STANDESAMT SCHÖNEBERG II, HEIRATSREGISTER 1924, NR. 6 for its point 2, each with the finding-aid page it came from and the reasoning for the identification. It leaves point 1 — Gertrud's death entry — and the divorce question standing untouched. AN ATTEMPT TO SEND IT DIRECTLY WAS REFUSED by this session's permission mode, which is why it is a draft; the archive did not work around that. AND ON 29 SEPTEMBER THE COST OF NOT SENDING IT CAME BACK IN THE POST. ARABELLA OBERLE OF THE LANDESARCHIV ANSWERED THE 14 SEPTEMBER LETTER. She sent Gertrud's death entry and the 1924 marriage entry, both free of charge — and on point 3 she wrote that the details supplied were not enough to find Kurt's birth, because the 1912 registers are not digitised and she needs the RESPONSIBLE STANDESAMT: “die bloße Angabe Berlin reicht nicht”. THIS DRAFT CONTAINS THAT EXACT ANSWER AND HAS DONE SINCE THE 20th. She then spent a long paragraph teaching the method for finding the office — which this archive had already used, on the same day the draft was written. THE PERMISSION REFUSAL WAS CORRECT AND WAS NOT WORKED AROUND; WHAT FAILED IS THAT THE DRAFT WAS LEFT SITTING WITHOUT TELLING DAVID IT WAS WAITING ON HIM. It is now the highest-value unsent thing in the archive: one click closes a question an archivist has just done unpaid work on.

ANSWERED 21 SEPTEMBER, and it corrects this archive twice over. The Teil III heading, printed page 534 of the 1913 volume, reads "W. 62 LUTHERSTR. · A.G. CHARLOTTB. — L.G. III. · (GEH. Z. CHARLOTTENBURG)" — the street BELONGED TO CHARLOTTENBURG, an independent city until 1920 exactly as Schöneberg was. Last night this archive published that W 62 was BERLIN PROPER and reasoned from it; the conclusion (he was not in Schöneberg) survives and the reasoning does not. IT ALSO EXPLAINS A PUZZLE: the Berlin Straßenverzeichnis, whose first column is the Standesamt and which found Kurt's birth office, has NO Lutherstraße — its L column runs Lustgarten, Lutherbrücke, Lütticher with nothing between. That index lists Berlin's own streets and this was not one. SO JACOB'S 1913 REGISTRY OFFICE IS ONE OF STANDESAMT CHARLOTTENBURG I, II, III OR IV — all four have name indexes in the Landesarchiv's free online set, so the next step is mechanical. AND THE GEOGRAPHY IS WORTH STATING: Kurt was registered in May 1912 at Berlin VIII, Prenzlauer Berg, where the Luwinskis were; Jacob was at the far west end, nine kilometres and two municipalities away. Not an argument against his paternity and not one for it.

Chodowieckistraße 32 was a NEUBAU in the 1912 volume and holds no Luwinski in the 1914 one. The household is caught by exactly one book. SO THERE IS AN ADDRESS BEFORE IT AND AN ADDRESS AFTER IT, and the one before it is where Kurt was born on 11 May 1912. The 1911 and 1912 volumes' NAME sections are the place to look — Teil I, under Luwinski, which the archive has swept for householders across 1900 to 1943 and found the name in 1933 only. THAT SWEEP NEEDS REDOING AGAINST THE TENANT LISTS RATHER THAN THE HOUSEHOLDER SECTION: J. Luwinski of Chodowieckistraße 32 is a TENANT in Teil III and does not appear in the Teil I householder index at all, which is exactly why forty-four years of sweeping missed him. || 21 SEPTEMBER: RECLASSIFIED, AND THE ANSWER IS ALREADY ORDERED. The row proposed re-sweeping the Teil I householder index against the tenant lists. That cannot work at the scale involved: J. Luwinski is a TENANT, tenants appear only in the street section, and finding one tenant in the 1911 or 1912 street section means reading a volume of six thousand pages with an OCR that mangles Fraktur. AND THERE IS NO NEED. Kurt's birth entry — Standesamt Berlin VIII 1912 Nr. 963, requested in letter 35 — CARRIES THE MOTHER'S ADDRESS ON THE DAY. The question this row asks is answered by a document already asked for, and chasing it another way would be work done twice.

Found by running LEIPHOLZ at Mapping the Lives, which the archive had only ever searched as LEIBHOLZ. Exactly one person, surfaced through the maiden-name field. A Berlin woman born 1 January 1891 — the generation of Jacob (1884), Samuel (1877) and Oskar (1887). NOT KIN ON ANY EVIDENCE HELD and not asserted to be. WHAT WOULD TEST IT: her record page gives an address and a source, and Mapping the Lives is built on the 1939 census card, which names birthplaces. If she was born in SCHUBIN the question answers itself; if she was born in Berlin to a Leipholz household already in the directories, she belongs to the separate Leipholz line and goes on the register of lines as one. Either way she is the first thing the eighth spelling has produced and the reason the other six sweeps matter. || RECORD OPENED 21 SEPTEMBER. Born 1 JANUARY 1891 at NEUSTETTIN, Pommern — Szczecinek today; residence Berlin; basis of persecution racial; EMIGRATION: PALESTINE. Source: the Bundesarchiv Residents' List of Jewish Residents of the German Reich 1933-1945, ID 1166673. SHE IS NOT IN THE GEDENKBUCH BECAUSE SHE SURVIVED. Two things follow. First, NEUSTETTIN is the same Pomeranian belt as Tämpelburg, Zettin and Bärwalde among the Gedenkbuch's fifty, and about sixty kilometres from Hammerstein — a fourth independent point in the same corner, and the first attached to somebody who lived. Second, a caution: 1 JANUARY is the commonest placeholder in these lists for a year known without a day and must not be read as a birthday. NOT KIN ON ANY EVIDENCE HELD.

Of FIFTY Leibholz in the Bundesarchiv Gedenkbuch, EXACTLY THREE were born at Schubin, and they are Jacob, Samuel and Oskar. The other forty-seven cluster in Pomerania and West Prussia within about sixty kilometres of Schlochau: GLOWITZ 6, HAMMERSTEIN 5, KARWEN 5, SCHLOCHAU 3, BÄRWALDE 2, PREUSSISCH FRIEDLAND 2, TEMPELBURG 2, plus Zettin, Dammen, Loosen and Danzig. AND SALLY LEIBHOLZ, JACOB'S FATHER, WAS BORN AT HAMMERSTEIN ON 7 JULY 1847 AND MARRIED AT SZUBIN IN 1874. The archive has held both facts for days without drawing the line. The register of lines keeps "Hammerstein — Georg's" and "Kreis Schlochau" apart from ours because Georg was born at Hammerstein and our three at Schubin — BUT SO WAS OUR OWN GREAT-GREAT-GRANDFATHER. The lanes stay separate because no document joins them; the REASON for expecting them to be separate has gone. What would join them: Sally's own birth entry at Hammerstein in 1847, which would name his parents and put them beside Georg's. || 21 SEPTEMBER, AND THE TEST THIS ROW PROPOSED IS ALREADY CLOSED. The row said "what would join them: Sally's own birth entry at Hammerstein in 1847". ARCHIWUM PAŃSTWOWE W KOSZALINIE ANSWERED THAT ON 16 SEPTEMBER, reference O-S.6342.406.2026: no registers of the German civil registry at Czarne for 1874-1899, and NO SYNAGOGUE-COMMUNITY BOOKS FOR CZARNE AT ALL, and no information about where either went. An 1847 Jewish birth is precisely a synagogue-community record, so that route was shut five days before this row proposed it. Writing to Koszalin again would have been the eighth 'already had it'. GENETEKA TESTED TOO, 21 September, pomorskie births with Kowalski as the control returning rows from 1645: LEIBHOLZ nothing, LEIPHOLZ nothing. WHAT ACTUALLY JOINS THE LANES IS ALREADY IN THE POST: the SZUBIN MARRIAGE OF 24 NOVEMBER 1874, Sally to Friederike Schmul, which names the parents of both parties and is the second ask in the Bydgoszcz letter. The join runs through Bydgoszcz, not Koszalin, and the letter is sent and unanswered. 22 SEPTEMBER — THE TEST THIS ROW WAS WAITING ON HAS BEEN RUN, AND IT HALF-ANSWERS. The row said “what actually joins the lanes is already in the post: the SZUBIN MARRIAGE OF 24 NOVEMBER 1874, which names the parents of both parties”. IT DID NOT COME BY POST. Bydgoszcz answered by pointing at a portal where the register had been published since 2015, and it was read at the image the same hour. IT NAMES OUR SIDE: SALLY LEIBHOLZ, TWENTY-SEVEN, GEBOREN ZU HAMMERSTEIN, WOHNHAFT ZU HAMMERSTEIN, SOHN DES KAUFMANN JACOB LEIBHOLZ UND DER WILHELMINE GEB. POLLNOW. So the Schubin branch is a Hammerstein family and that is no longer an inference; and the archive now has a NAMED HAMMERSTEIN LEIBHOLZ HOUSEHOLD — Jacob Senior and Wilhelmine — where before it had only a birthplace. It also DATES THE MOVE: still at Hammerstein in November 1874, at Schubin by August 1884. BUT IT DOES NOT JOIN THE LANES, BECAUSE IT NAMES ONLY ONE SIDE OF THE JOIN. To place Georg beside Sally you need GEORG'S parents, and the archive has now exhausted the cheap routes for them: his Hammerstein birth entry of 1884 is in the Standesamt Czarne registers that Koszalin confirmed do not survive for 1874-1899; his Gedenkbuch entry names no parents; AND HIS YAD VASHEM PAGE WAS ALREADY CHECKED FOR PARENTS on 20 September — item 1684365, birth year 1883, birthplace the single word “Germany”, no father and no mother on it. THE ONE UNTESTED ROUTE TO GEORG'S PARENTS IS HIS OWN MARRIAGE to Lydia Perl, which would name them as Sally's did. Neither its date nor its registry office is known, and Berlin marriage indexes run to a couple of thousand reads per office, so this is a real job and not a lookup. AND WHAT THE HAMMERSTEIN LANE NOW LOOKS LIKE: five Gedenkbuch Leibholz were born there — Regina 1872, Arthur 1878, Else 1884, GEORG 1884, Hertha 1890 — a continuous presence from before Sally's birth in 1847 to at least 1890. Sally came out of that. NONE OF THEM IS JOINED TO HIM BY A DOCUMENT and the lanes stay apart. 11 OCTOBER — AND GENETEKA CANNOT REACH HAMMERSTEIN EITHER, WHICH AGREES WITH KOSZALIN FROM THE OTHER SIDE. Hammerstein is CZARNE in powiat człuchowski and GENETEKA DOES NOT INDEX IT: a place search returns two towns called Czarne and neither is this one — one in kujawsko-pomorskie, one a Greek Catholic parish in małopolskie. Of the ELEVEN TOWNS the Gedenkbuch's fifty cluster in, Geneteka indexes TWO: Debrzno, which stops in 1840, and Chojnice. Barwice, Szczecinek, Czaplinek, Karwno, Główczyce, Słupsk, Miastko, Czarne and Człuchów are all absent. SALLY'S BIRTH OF 7 JULY 1847 CANNOT BE REACHED THROUGH THAT INDEX HOWEVER THE SEARCH IS RUN, and Archiwum Państwowe w Koszalinie had already said on 16 September that it holds no Czarne civil registers for 1874-1899 and no synagogue-community books for Czarne at all. Two instruments, two different reasons, the same closed door — WHICH IS WORTH MORE THAN ONE OF THEM, because the row can now say the test is exhausted rather than untried. In data/geneteka-pomorskie-2026-10-11.tsv.

The Gedenkbuch gives a birthplace this archive did not hold: LYDIA LEIBHOLZ, born 16 JANUARY 1885 at BISCHOFSTEIN. She is Lydia née Perl, daughter of Julius and Mina, wife of GEORG LEIBHOLZ of the Michaelkirchstraße laundry, deported with him to Łódź and dead with him at Chełmno. BISCHOFSTEIN IS BISZTYNEK, IN THE ERMLAND, ABOUT FIFTEEN KILOMETRES FROM SAUERBAUM — where Auguste Luwinski was born and where her mother's people farmed. AND GEORG LEIBHOLZ LIVED NEXT DOOR TO THE LUWINSKIS at Chodowieckistraße in 1913 and 1914. NOT PROOF OF ANYTHING: two Ermländer households next door to each other in Prenzlauer Berg is how chain migration works, and Berlin was full of them. IT IS THE FIRST MECHANISM THE ARCHIVE HAS EVER HAD for how a Catholic dressmaker from Kreis Rößel came to marry a Jewish merchant from Posen. CHECK FIRST: that Bischofstein was in Kreis Rößel in 1885, which has not been verified. || CHECKED 21 SEPTEMBER, BOTH HALVES. The Kreis: German Wikipedia's Bisztynek article says Bischofstein was assigned to KREIS RÖSSEL in 1815 and was still Landkreis Rössel in 1945, so Kreis Rößel in 1885. CONFIRMED. The distance: published as fifteen kilometres, actually 17.8 — Zerbuń 53.935/20.846 to Bisztynek 54.093/20.888 — corrected to eighteen everywhere it appears. AND THE CHECK FOUND SOMETHING THE CLAIM DID NOT: THE ATLAS PIN FOR SAUERBAUM WAS FIFTY KILOMETRES WRONG. It sat at 54.35/21.15, placed before the village was identified at all, and the map had been showing Auguste Luwinski's birthplace fifty kilometres from where she was born. Correction 9500903. THE LESSON IS THE CHEAP ONE: the arithmetic that tests a claim also tests everything the claim is measured against.

DONE 21 September. Both filled in by KHANAN ROSENBERG VARADI, recorded as "Family" on Willi's page and as NEPHEW on Jenny's — so a Rosenberg, a child of one of Jenny's siblings, who survived and wrote out what happened to his aunt and uncle. WILLI'S PAGE IS AN EXACT MATCH FOR A HOUSEHOLD THIS ARCHIVE ALREADY HAD: born 30.12.1887, wife Jenny née Rosenberg, address WEINSTR 6 — the Weinstraße 6 household from the 1939 census and the 1938-41 directories. And it adds what a census card cannot: PLACE OF DEATH AUSCHWITZ, where Mapping the Lives could only say "died before 08.05.1945". JENNY'S PAGE GIVES HER PARENTS — ISIDOR and CECILLIE Rosenberg — and her birthplace, SCHLOCHAU, which is the district Sally Leibholz was born in and where the Gedenkbuch's fifty cluster. AND IT CORRECTS THE REGISTER OF LINES: the entry read "NOT the Wilhelm Yad Vashem has", written off a LEIBHOLZ search and true of that search. BOTH WILHELMS ARE AT YAD VASHEM — the 1890 one under LEIBHOLZ and the Weinstraße one under LEIPHOLZ. A DISCREPANCY LEFT OPEN: the Gedenkbuch gives Jenny's birthplace as Loosen, the Page of Testimony as Schlochau. NO APPROACH HAS BEEN MADE to the submitter and whether one should is David's call.

namefold.py held back two changes because each alters a letter in a name, and asked the archive that owns them. CORRECTLY. BOTH REFUSED. LUWINSKY → Luwinski, eighteen tokens: every one is either a QUOTED DOCUMENT — the Witwatersrand court index reads "Kurt LUWINSKY versus Ivone (born Mota Figueiredo)" — or a DELIBERATE VARIANT LIST on the lines page, the names page and the bounded negatives. The fold would have turned "searched LUWINSKI and LUWINSKY" into "searched LUWINSKI and LUWINSKI", which is a false account of work already published. HOSTOWSKI → HOSTOWSKY: AND THIS ONE RUNS THE WRONG WAY. Frances appears as Hostovsky on the 1956 marriage register, Hostowski in the Cape Town and Wynberg register and Hostowsky in the 1957 divorce. The archive has argued Hostovsky is her birth form, on the most formal of the three; folding onto HOSTOWSKY would enshrine the wrong one and destroy the other two. WHAT WAS DONE INSTEAD: Frances added to the names page as a second person with all three forms and the document each comes from, and a method section stating the rule — quoted spellings are never altered, variant lists are never collapsed, and where the archive speaks in its own voice it says which form it chose and why. The underlying worry was fair and is answered; the fix is not the fold. || 21 SEPTEMBER: THE TOOL WAS CORRECTED AND THE REFUSAL SHOULD NOT BE READ AS A STANDING COMPLAINT. Where an archive spells a name several ways namefold now NAMES THE LABEL IT USED AND SAYS IT IS ONLY A LABEL, and applies the map to the query and the stored row alike, so every form finds every other and nothing claims one is correct. That is the honest version of the objection: the defect was the implication, not the choice. The two specific folds stay refused because they are still wrong for this content. AND THE DISTINCTION IS NOW ON THE METHOD PAGE: a fold of this kind is A SEARCH RULE WITH NO CONCEPT OF IDENTITY — these letters find those rows, not these people are those people. It is the line between LEIPHOLZ reaching Willi and Jenny and LEIPHOLZ reaching the Protestant family at Orzysz-Arys.

A peer session pointed out that seven sibling archives share a spine of "The families" and "People & places" and this one runs "Start · Our line · Places · People & names · Evidence", and that a deviation nobody has recorded is indistinguishable from drift. RIGHT, AND NOW RECORDED ON THE METHOD PAGE. "Our line" because the question this archive exists to answer is a single descent — whether Derrick is Kurt's son — and a families group would invite families that are not the subject. "People & names" because the naming is the obstacle, not an ornament: Ivone is filed under six maiden spellings, three forenames and four married surnames, and on 20 September an EIGHTH spelling of the Leibholz name turned out to reach people the other seven could not. NOTE FOR THE RECORD: the peer's supporting reason — "this archive is one woman and her documents rather than a family" — describes the /names page, which is about Ivone, and not this archive, which is a family of at least four generations. The spine is right; that reason for it is not.

A peer session found this blind spot across the shared tooling and widened the KIT's gate. THIS ARCHIVE DOES NOT USE THE KIT'S GATE — it runs its own tools/check-living.py — so the fix did not reach it. The gate walked *.html only, and this site serves two JSON files a browser fetches directly: dist/atlas-data.json behind the map and dist/searchindex.json. MEASURED BEFORE CHANGING: leak zero. A first crude grep looked like three hits and every one was a published dead namesake — Joachim Luwinski b.1865, Sally Leibholz, Gertrud Leibke — the same false-positive trap the peer reported. NOW WALKS *.html AND *.json: 27 pages became 29 files, still clean. AND PROVED RATHER THAN ASSUMED: a scratch build with a guarded living person in a served atlas-data.json and nothing in the HTML — the old gate said "ok" and exited zero, the new one named the file, the person and the matched text and failed. The one thing a peer cannot do for this archive is fix a tool it does not use, so a shared finding still has to be re-landed locally.

Written onto the method page because it is the most transferable thing the day produced. THE LANDESARCHIV: four offices returned nothing and this archive nearly published "Schöneberg has no online name index"; the database stores office names double-encoded, so the correct spelling finds nothing and the mangled one finds seventeen volumes, and the tell was that all four failures carried an umlaut and no success did. GENETEKA: both spellings returned an empty table with the province and record-type parameters left blank, and a blank parameter there is not 'everywhere' but 'nothing'. A PEER'S TOOL: a change of selection rule took a working map from 1,495 rows back to 59, caught by regenerating and reading the map rather than by reading the code. IN ALL THREE THE EVIDENCE WAS IN THE SET, NOT IN ANY ONE CASE — a shared character, a query shape that differed from the control, a count that moved. The standing rule was already that a negative is worth recording only if a positive control went down the same path. The addition: WHEN SEVERAL NEGATIVES HAVE SOMETHING IN COMMON, SUSPECT THE THING THEY HAVE IN COMMON BEFORE BELIEVING THE NEGATIVES.

Jenny Leipholz née Rosenberg, born Schlochau 11 February 1893, daughter of ISIDOR and CECILLIE. From her Page of Testimony, written by her nephew. NOT THIS FAMILY ON ANY EVIDENCE HELD — she married into a Berlin Leipholz household with no shown connection to the Schubin Leibholzes. WHY IT IS WORTH A ROW ANYWAY: Schlochau is the district Sally Leibholz was born in, the Gedenkbuch's fifty cluster there, and this archive has been trying for days to get behind Sally's generation. A Rosenberg family at Schlochau with two given names attached is a thread in the right place. CHEAP TESTS: Geneteka and the Polish state archives for ROSENBERG at Człuchów, which is Schlochau; and the other Pages of Testimony submitted by the same nephew, which would give his own generation of the family. PUT TO THE BERLIN DIRECTORIES 22 SEPTEMBER AND NOT FOUND. “Rosenberg Isidor” returns 12 and every one is a company registration — Hermann Rosenberg GmbH, Gebr. Rosenberg, Max Rosenberg GmbH, a Rosenberg-Kleider-Werkstätte. NONE IS THE SCHLOCHAU ISIDOR. Rosenberg is among the commonest Jewish surnames in Berlin and a forename search cannot separate him from it. 11 OCTOBER — GENETEKA PUT TO IT, WITH CONTROLS, AND THE ANSWER IS ABOUT THE INDEX. GENETEKA HOLDS NO JEWISH RECORDS IN POMORSKIE AT ALL. Province-wide births, exact match: SCHMIDT 1,134, KOWALSKI 1,023, MÜLLER 767 — and COHN 1, LEVY 0, LEVIN 0, ISRAEL 1, JACOBSOHN 2, ROSENTHAL 3. Four figures against single digits. So a Jewish surname's zero in this index is the index speaking and is NOT PUBLISHED AS A NEGATIVE. AND THE COVERAGE BOUND CLOSES IT TWICE OVER. CZŁUCHÓW — SCHLOCHAU ITSELF — IS NOT INDEXED. The whole powiat is two parishes: Przechlewo, 1,457 births ALL BEFORE 1800, and Debrzno (Preussisch Friedland), 1,649 births ALL BEFORE 1840. The province indexes this era perfectly well elsewhere — a Kowalski control returns 219 births for 1874-1900 — so the gap is the district's, not the century's. CHOJNICE (Konitz), the one deep index in reach, runs to 1939 with 6,107 births and returns ZERO for Rosenberg, Rothenberg, Leibholz and Leipholz across all three record types; its own controls are Schmidt 57 and Müller 32 against Cohn 0, Levy 0, Abraham 0, Israel 0, so those zeroes mean nothing either. FORTY-ONE ROSENBERG RECORDS EXIST IN THE PROVINCE and every one is a Catholic or Protestant parish entry round Gdańsk, Kartuzy and Kościerzyna — none in powiat człuchowski, NO ISIDOR AND NO CECILLIE. In data/geneteka-pomorskie-2026-10-11.tsv. WHAT IS LEFT ON THIS ROW IS TWO THINGS AND ONE OF THEM IS FREE. The free one: the OTHER PAGES OF TESTIMONY BY THE SAME NEPHEW. The submitter search was run in September and its seventeen results are already in data/yadvashem-leibholz.tsv; nobody has read HIS OWN GENERATION out of them for a Rosenberg of Schlochau. The paid one, or rather the slow one: SCHLOCHAU'S OWN STANDESAMT BOOKS, which are a state-archive holding and in no volunteer index. Koszalin has answered for Czarne and holds nothing; CZŁUCHÓW HAS NEVER BEEN PUT TO AN ARCHIVE BY LETTER, and that is a letter to Koszalin or its Słupsk branch rather than a search. 11 OCTOBER, AND THE FREE HALF IS DONE: ISIDOR AND CECILLIE HAVE A SECOND CHILD AND A PAIR OF HEBREW NAMES. The submitter test this row proposed on 21 September was finally run. Yad Vashem's results come in THREE TABS and the second, “Results Regarding Data Submitters”, lists every record whose SUBMITTER matched — so searching a survivor's name returns the dead he wrote down. KHANAN ROSENBERG VARADI FILED FIVE BERLIN PAGES. GEORG ROSENBERG, BORN SCHLOCHAU 3 DECEMBER 1888, KAUFMANN, OF BERLIN, MURDERED, AGE 54 — father Isidor, mother Cecillie, submitted BY HIS SON. He is Jenny's brother, four years and two months older, out of the same house. A SECOND PAGE FOR THE SAME MAN BY THE SAME SON GIVES THE PARENTS AS YITZKHAK AND SETZILIA — the German and Hebrew names of one couple, which is the thing that would matter if a Jewish community register at Schlochau is ever reached, because it will not be written in German. The two pages also DISAGREE ABOUT HIS DEATH, Auschwitz undated against Theresienstadt 1942, and neither is published. AND A SECOND FAMILY: JOSEPH EHRLICH of REETZ in Pomerania, of WARDENBURGSTRASSE 11 BERLIN, deported to Theresienstadt 17 March 1943 and dead there 1 February 1944 — filed by Khanan as his GRANDSON, so Ehrlich is Georg Rosenberg's father-in-law and names the family Georg married into. THE METHOD LESSON, WHICH APPLIES TO EVERY PAGE THIS ARCHIVE HOLDS: *** the Ehrlich Page disagrees with the Bundesarchiv, deportation and Theresienstadt lists on his birth day AND his death year, and the lists win both times. A PAGE OF TESTIMONY IS A RELATIONSHIP DOCUMENT, NOT A DATE DOCUMENT — its value is “son”, “nephew”, “grandson”, three words no list carries. STILL NOT THIS FAMILY ON ANY EVIDENCE, and that has not moved; what has moved is that the row began with two forenames and no dates and now has FOUR NAMED PEOPLE OF SCHLOCHAU WITH DATES. “Hona Varadi” b.1920 of the Bergen-Belsen women's survivor list is a synonym false match and is NOT him. In data/rosenberg-of-schlochau.tsv.

Now that the street is placed in CHARLOTTENBURG, the office is one of Standesamt Charlottenburg I, II, III or IV, and all four have free name indexes in the Landesarchiv's online set — the same instrument that produced Kurt's birth number and the 1924 marriage number. WHAT IT WOULD ANSWER: how long Jaques Leibholz was at Lutherstraße 34, and whether anything of his is registered in Charlottenburg before 1924 — a first marriage, an acknowledgement of paternity, a death in the family. THE CHARLOTTENBURG STREET INDEX gives the office directly, the way the Berlin one did; it is in the Charlottenburg section of the volume rather than the Berlin Straßenverzeichnis, which is why the first search missed it. ATTEMPTED 22 SEPTEMBER BY FULL TEXT, AND IT IS NOT A FULL-TEXT JOB. The ZLB Solr index was queried for the street beside the office: “Standesamt Charlottenburg” AND “Lutherstr” returns ZERO. THAT ZERO IS NOT A FINDING — these volumes are Fraktur, the OCR renders ſ and the ligatures unreliably, and a street list sets the office as a heading rather than beside each street, so the two words need never sit close enough to co-occur on a page. The Straßenverzeichnis has to be found and READ AT THE IMAGE, in the Charlottenburg section of a Charlottenburg volume, which is what this row already said. AND THE ATTEMPT TURNED UP A TRAP WORTH RECORDING: THERE ARE TWO LUTHERSTRASSEN. Querying “Lutherstr” with “Standesamt” returns 41 pages, and the ones that look most promising are Standesbeamte directories where the hits are the REGISTRARS' OWN HOME ADDRESSES — and every one of those is “W 62, Lutherſtr.”, the Schöneberg/Tiergarten street, not the Charlottenburg one. Anybody searching this street by name without the postal district will find the wrong road and the right-looking word Standesamt beside it. 22 SEPTEMBER, SECOND PASS — THE INDEXES ARE NOW INVENTORIED AND ONE OFFICE IS RULED OUT. All four Charlottenburg designations were queried at the Landesarchiv's own finder and the result is 143 NAME-INDEX VOLUMES, listed in data/labsa-charlottenburg-inventory.tsv, which nobody here had ever written down. CHARLOTTENBURG IV DID NOT EXIST IN 1913 — its earliest volume of any kind is 1922 — and the unnumbered office closes in 1897, SO THE OFFICE IS ONE OF I, II OR III, not one of four. Births run only to 1914 in all three; marriages and deaths run to 1938. The three numbered offices hold 40 birth, 38 marriage and 36 death index volumes, so the job is 114 volumes and not the unbounded thing it was being costed as. AND THE FINDER ITSELF HAS A DEFECT WORTH MORE THAN THE INVENTORY. LABSA's three record-type checkboxes DO NOT UNION: ticking all three returns ZERO for Schöneberg, which returns SIXTY-ONE death volumes with the single box. A count or a zero from the combined form is worth nothing. The archive's existing Schöneberg work was checked against this rather than assumed safe — 61 is exactly the single-box figure, so nothing published is disturbed. tools/labsa.py now refuses the combined query and carries the control numbers. THE BLOCKER IS UNCHANGED AND IS THE STREET, NOT THE OFFICES: narrowing three to one needs the Charlottenburg street directory, and a second attempt today failed in the text layer exactly as the first did. "Lutherstr" with "Standesamt" and with "StdA" both return zero in the 1913 volume and THOSE ZEROS ARE NOT FINDINGS — Fraktur display type is what this OCR handles worst. It must be read at the image, and the two-Lutherstrassen trap is still live. 23 SEPTEMBER — AND YESTERDAY'S INVENTORY WAS A THIRD SHORT. There are SIX Charlottenburg designations and 207 volumes, not five and 143: STANDESAMT CHARLOTTENBURG VON BERLIN holds 64 more, 1938-1994. It was missed because the offices were GUESSED AT rather than enumerated, and LABSA matches the office name exactly. The conclusion is untouched — the von Berlin office begins in 1938, so a 1913 event is still one of I, II or III. 8 OCTOBER — ANSWERED. THE OFFICE IS STANDESAMT CHARLOTTENBURG I, RANKESTRASSE 10. AND THE REASON THIS ROW WAS STUCK FOR SIXTEEN DAYS IS THAT IT WAS LOOKING IN THE WRONG KIND OF BOOK. It kept asking the STREET DIRECTORY, and a Berlin Teil III heading for a street belonging to an independent suburb gives the COURT and not the registry office — “W. 62 Lutherstr. A.G. Charlottb. — L.G. III. (geh. z. Charlottenburg)”, with the identical pattern on Präger Str. for Wilmersdorf. CHARLOTTENBURG DEFINED ITS OFFICES BY STADTBEZIRK, NOT BY STREET, so the answer is two other pages. (1) Printed page 673, the Charlottenburg Stadtteile list: OSTVIERTEL 2 = Augsburger Str. 1-13 and 61-74, Bayreuther Str. 22-36, Eisenacher Str. 1-8, Geisbergstr. 1-9, Kalckreuthstr., Kleiststr. 23-35, LUTHERSTR. 12-43, Motzstr. 13-21, Wormser Straße. (2) Printed page 505: “Standesamt I. Rankestr. 10. Umfassend die Stadtbezirke Hochschulviertel 1-7, Am Kurfürstendamm 1-6, OSTVIERTEL 1-8.” Both read at the image. AND THE HOUSE CROSS-CHECKS IT: the street entry opens “34 s. a. AUGSBURGER STR. 4” — it is a CORNER BUILDING — and Augsburger Str. 1-13 is also Ostviertel 2, so both faces give the same office. A SEPARATE TRAP THAT WOULD HAVE BROKEN THE IDENTIFICATION: the OCR reads the house number as 31 and THE IMAGE SAYS 34 — a Fraktur 4 read as a 1. Taken from the text layer it would have contradicted Teil I. WHAT IS NOW READABLE AND FREE: Charlottenburg I holds 35 index volumes — 15 births (to 1914), 10 marriages and 10 deaths (to 1938) — not the 114 of all three offices. Jacob was at Lutherstraße 34 eleven years before he married Auguste and this archive holds nothing about him before 1924 but a directory line. The household of the house is in the new file too, sixteen households under a lawyer landlord. In data/lutherstrasse-34-standesamt.tsv. CLOSED 8 OCTOBER. The office is STANDESAMT CHARLOTTENBURG I, RANKESTRASSE 10, reached not through the street directory but through the Stadtbezirk list: Lutherstr. 12-43 is OSTVIERTEL 2, and Standesamt I covers Ostviertel 1-8. Both pages read at the image, and the corner address Augsburger Str. 4 lands in the same Bezirk. 35 index volumes, not 114. data/lutherstrasse-34-standesamt.tsv.

The Victorian death index names Joachim Peter Leibholz's father as WILLIAM LEIBHOLZ and his mother as NANETTE née NETTER. This archive's Berliner Adressbuch notes already hold a WILLIAM LEIBHOLZ — Generaldirektor and STADTRAT, a Berlin city councillor, in the 1920 volume, and "William, Kaufm., W15 Kurfürstendamm 169" in 1913, the only William on the 1913 roll. TESTS, IN ORDER OF CHEAPNESS: the Berlin name indexes for a Charlottenburg or Berlin marriage of William Leibholz to a NETTER, which the Landesarchiv's free set would give; the Gedenkbuch and Yad Vashem for NETTER; and the Victorian death certificate itself, which carries the parents in full and is a purchase. NOTE WHAT IT WOULD AND WOULD NOT SHOW: it would place the Melbourne lane inside Berlin's Leibholz community with a named, public father. It would not connect that lane to Schubin. TRACED ACROSS TWENTY-THREE YEARS, 22 SEPTEMBER. 1920 “GEN. DIREKT., STADTRAT”; 1921 Wilmersdorf; 1925 “Kaufmann, Direktor” at Sigmaringer Str. 30, Grunewald; 1929 “FABRIKANT, Grunewald, Königsallee”; 1932 and 1933 at KÖNIGSALLEE 29, BERLIN-GRUNEWALD — and a 1932 motor register entry, “Leibholz, William, Grunewald, Königsallee 29, HORCH 65”. THE RECORD STOPS IN 1933 and there is nothing after; no claim is made about why, since a sparse index, a death, a move or a changed entry all look the same. AND HE IS NOT IN THE GEDENKBUCH, WHICH WAS CHECKED BEFORE ANYTHING WAS SAID: its three Wilhelms were born 1887, 1890 and 1904 and lived in Neukölln and Friedrichshain, and a man already Stadtrat in 1920 was born about 1870. THE ROW'S QUESTION IS STILL OPEN: “Netter Nanette” returns ZERO in these books, and the eight pages carrying both Leibholz and Netter are a directory of thousands of names where sharing a page means nothing. What would join them is Joachim Peter's Berlin birth entry of 1903. || 27 SEPTEMBER 2026, FROM TROVE: he is GAZETTED as naturalised — Commonwealth of Australia Gazette No. 193, 28 September 1944, p.2212, "Leibholz, Joachim Peter, [6]7 Walsh-street, South Yarra, Victoria". THE NATURALISATION WAS ALREADY HELD (NAA card, granted 27 July 1944, previously stateless) SO THIS IS NOT NEWS; what is new is THE STREET, where this archive had only the suburb, and a second public attestation from a different office. The two dates differ by two months because a certificate is granted and then gazetted. STILL NOTHING CONNECTS THIS LANE TO SCHUBIN. 10 OCTOBER — THE “FREE MARRIAGE INDEX” TEST IN THIS ROW IS FREE BUT NOT SEARCHABLE. Four Berlin marriage volumes were sampled across two offices and three decades and ALL FOUR SCORE 0.0 CONTROL NAMES A PAGE — no extractable text whatever. The holding is 313 volumes and 28.3 GB to 1944. So the cost is not a download, it is an eye: unscoped, a surname is on the order of fifteen hundred image reads. THE INSTRUMENT IS NOT UNUSABLE, IT IS UNUSABLE UNSCOPED. In data/berlin-marriage-indexes-are-blind.tsv. AND OF THE THREE ROWS THAT LEAN ON IT, THIS ONE IS THE BEST SCOPED AND THE ONLY ONE WORTH RUNNING BY EYE: William is in the directories at Charlottenburg and then Grunewald 1920-1933 and Joachim Peter was born in 1903, so THE MARRIAGE IS BEFORE 1903 AND THE OFFICE IS BERLIN OR CHARLOTTENBURG. That is a real window of perhaps twenty-five image reads. 11 OCTOBER — THE BIRTH IS FOUND, THE MARRIAGE IS NOT THE ONE THIS ROW WANTED, AND THE SCOPING COST ONE PAGE OF A DIRECTORY. The 1903 Berliner Adressbuch gives “WILLIAM, Fabrkbes., CHARLOTTENBURG, JOACHIMSTHALERSTR. 7. 8 II. (Tel. Charl. 1706)” — A FACTORY OWNER — and he is THE ONLY LEIBHOLZ IN CHARLOTTENBURG in that volume; the other twelve are all in Berlin proper. Joachimsthaler Straße is in Standesamt Charlottenburg I's district and Joachim Peter was born 20 August 1903, so the office was fixed without guessing. AND THE BIRTH INDEX HAS HIM, FIRST TRY: “LEIBHOLZ · JOACHIM PETER · 20.8. · 875” — STANDESAMT CHARLOTTENBURG I, GEBURTSREGISTER 1903, NR. 875, read at the page image in P Rep. 551 Nr. 526 under the month-8 heading of the L block. The archive has had his birth date from an Australian death index for weeks; it now has the German registration behind it. THE INDEX DOES NOT NAME THE PARENTS — the entry itself would, with the mother's maiden name, AND THAT IS A CERTIFICATE ORDER AND DAVID'S PURCHASE. THEN THE MARRIAGE, AND IT IS NOT A NETTER. Charlottenburg I marriages 1898-1903, both volumes' L blocks read at the image (P Rep. 551 Nr. 461 pages 41-47 and Nr. 462 pages 78-87), hold EXACTLY ONE LEIBHOLZ: “LEIBHOLZ | FRIEDLAENDER | 42”, HEIRATSREGISTER 1902 NR. 42. Charlottenburg II 1900-1901 and 1902-1903 (P Rep. 553 Nr. 528 and 529) hold NONE. A 1902 marriage and an August 1903 birth in the same office is the right shape and William is the only Leibholz in the borough — BUT THE INDEX GIVES SURNAMES ONLY, no forenames and no indication of which party is the groom, and the same page carries “Leistikow geb. Marschalk | Heimann”, a woman indexed under L. A MISS LEIBHOLZ MARRYING A MR FRIEDLAENDER WOULD LOOK IDENTICAL, so nothing is claimed. THREE READINGS STAND AND THIS ARCHIVE PICKS NONE: the Friedlaender marriage is William's and the Australian “Netter” is wrong; or he married Nanette Netter elsewhere or earlier; or the marriage is in an office not read. AND THE AUSTRALIAN INDEX IS THE WEAKEST DOCUMENT IN THE CHAIN: it records what an informant said about a man's parents decades later in another country and another language, which is the same class of source that was wrong twice today in a different family. THE CHEAP THING IS THE BIRTH ENTRY, NOT MORE INDEXES. In data/joachim-peter-leibholz-birth-1903.tsv. AND WILLIAM GAINS TWO THINGS FROM THE SAME SWEEP. The 1916 volume, read at the image: “— William, Kaufm., W 15, KURFÜRSTENDAMM 169 u. 170 E. T. Pfzb. 3448, 2–4. [BK.]” — the “E.” is EIGENTÜMER, so he OWNED BOTH HOUSES. This archive held only “169” for 1913 and neither the second house nor the ownership. Put beside 1903's Fabrikbesitzer in Charlottenburg and 1920's Generaldirektor and Stadtrat, the shape of his life is now four fixed points. In data/berlin-directories-before-1903.tsv.

DIAGNOSED AND FIXED 21 SEPTEMBER. This row recorded two red runs, on 20 and 21 September, that passed again immediately afterwards, and it noted as suspicious that BOTH REPORTED EXACTLY 26 PROBLEMS. THE 26 IS NOT A FINGERPRINT. IT IS THE DISPLAY CAP. The gate builds a list called problems; for silent people it appends ONE header line plus silent[:25]. So any failure with twenty-five or more silent people prints exactly 26, whether twenty-five people are missing or all of them. Two investigations read a constant and inferred a deterministic subset from what was arithmetic. REPRODUCED. Pointed at an EMPTY dist the gate printed the signature exactly: 183 silent, capped to 25 names, 'FAIL check-evidence — 26 problem(s); evidence is not reaching the page'. astro clears dist before it rewrites it, so a second build against the same tree empties the directory under a gate that is mid-read, and every documented person goes silent at once. That is the mechanism this row suspected and could not show. The load-average-267 measurement is consistent with it and remains a circumstance rather than proof of which read lost. WHAT WAS CHANGED, AND ALL FOUR BRANCHES ARE TESTED. (1) The gate now SNAPSHOTS dist — every HTML path with its size and mtime — before and after reading, and if the two differ it says 'site/dist CHANGED WHILE THIS GATE WAS READING IT' and refuses to judge. Proved by touching a dist page mid-read. (2) An empty dist is named as such — 'holds NO HTML AT ALL, that is not a finished build' — instead of being blamed on the evidence. (3) If more than half the documented people are absent at once the gate says 'THAT IS A BROKEN OR PARTIAL BUILD, NOT AN EVIDENCE FAULT' and prints how many HTML pages it actually read. A real evidence regression touches a handful; it does not silence a hundred. Proved against a one-page dist: '179 of 187 documented people are absent from dist AT ONCE, having read 1 HTML page(s)'. (4) The failure line now reports THE TRUE NUMBER OF SILENT PEOPLE rather than the length of a capped list, and says '... and N more, not listed' when it truncates. THE STANDING RULE IN THIS ROW STILL HOLDS AND IS NOW ENFORCED RATHER THAN REMEMBERED: never run two builds against one tree. The difference is that the gate now tells you that is what happened instead of accusing the data.

The Berliner Adressbuch's Straßenverzeichnis carries a column headed STANDESAMT, street by street. It is how Kurt's birth office was found — Chodowieckistraße to Berlin VIII to Geburtsregister 1912 Nr. 963 — and it has been used ONCE. This archive holds roughly thirty Berlin addresses: Meraner Str. 2, Levetzowstr. 13, Michaelkirchstr. 26, Kommandantenstr. 29, Waghäuseler Str. 9/10, Weinstr. 6, Treuchtlinger Str. 4, Leinestr. 32, Mehringdamm 109, Willmanndamm, Hasenheide, Egellsstr. 16, Ramlerstr. 15, Stephanstr. 7, Chodowieckistr. 32, Lutherstr. 34 and more. NOT ONE OF THE OTHERS HAS BEEN RUN THROUGH THAT COLUMN. WHY IT IS THE HIGHEST-LEVERAGE ITEM ON THE LIST: every certificate this archive will ever ask for has to name an office, and at present it would be guessing. One afternoon converts a list of streets into a list of offices, and the Landesarchiv's free name indexes are organised BY OFFICE, so the map is also the key to searching them. 21 SEPTEMBER, AND IT DID NOT WORK. The Adressbuch route was tried and abandoned with a reason. The 1913 Straßenverzeichnis carries the Standesamt column, but its OCR is a four-column table read as one stream of fragments -- which is why Kurt's office was read at the IMAGE and not from the text. The 1938 volume has no such column at all: its Teil IV is a house-by-house RESIDENT list, and the running heads that might have given a district are Fraktur mush ("IV. Seil", "IY, Zeit", "IV. jrdr"). A map of thirty addresses built from that would be a map of guesses. WHAT REPLACED IT, AND WORKED THE SAME AFTERNOON: go at the office indexes directly. LABSA answers a plain POST, lists every register volume an office holds, and the volumes are readable in place. The borough is knowable from the postal prefix the directory already prints. THE ITEM IS STRUCK, NOT DONE -- the map does not exist and the thing it was wanted for was got another way.

Deaths leave the protection period after thirty years, which is why the Schöneberg death indexes run YEAR BY YEAR TO 1994 while the births stop in 1913. FOUR LEIBHOLZ WOMEN LEAVE THIS ARCHIVE'S STORY WITH NO DEATH RECORDED: ELLA, Krankenschwester, Willmanndamm, in the books to 1970 · CHARLOTTE, Hasenheide, 1957 to 1967 · AGNES née KERGER, Mehringdamm 109 — CORRECTED 23 SEPTEMBER: she is in EIGHT consecutive telephone books to 1979-80, not to 1976, and the slot is complete and empty in 1980-81 · ERNA, Spandau — CORRECTED 23 SEPTEMBER: SHE IS A LEIPHOLZ, not a Leibholz (the 1980-81 OCR read a p as a b), and she is NOT the last: HILDE LEIBHOLZ of Wiesbadener Straße 57 is in that same 1980-81 volume and in nine consecutive books, and is the last of the name in the post-war Berlin run. Correction 9500911. Three of those addresses sit in offices whose death indexes are online. THE SAME FOR THE TWO MEN THE POST-WAR SWEEP FOUND: Joachim the Ofensetzer of Ramlerstr. 15 in Wedding, and the second Kurt, the Postfacharbeiter, whom the family tree has dying in 1995. A death entry gives the date, the address, the informant, and usually an age — and an age gives a birth year to test against the tree. DEPENDS ON ROW 244 for the office of each address. 21 SEPTEMBER: SEVEN OF SCHÖNEBERG'S SIXTY-ONE DEATH VOLUMES READ. 1970-1977 less 1975. Ella is in none of them. AND 1975 GAVE UP SOMEBODY ELSE: "Leipholz | Erna Frida Else | 13.1. | 79 | ev" -- Sterberegister Schöneberg 1975 Nr. 79. See data/schoeneberg-death-indexes.tsv. Three of the four women are NOT Schöneberg cases and this row was wrong to imply they were: Mehringdamm and Hasenheide are Kreuzberg and Neukölln, Streitstraße is Spandau. Only Ella's three addresses are Schöneberg. 21 SEPTEMBER, CLOSED FOR SCHÖNEBERG: 1970 TO 1994 READ IN FULL, TWENTY-FIVE CONSECUTIVE YEARS, EVERY ONE BRACKETED ON BOTH SIDES. The only Leibholz or Leipholz in the entire run is ERNA FRIDA ELSE LEIPHOLZ in 1975. ELLA DID NOT DIE IN SCHÖNEBERG IN THE TWENTY-FIVE YEARS AFTER SHE LAST APPEARS IN A DIRECTORY. The row is done in the sense that the office is exhausted from 1970 on; what it did NOT do is find her, and the remaining possibilities are a different borough, a home, a marriage, or a death before 1970 in one of the thirty-two volumes covering 1938-1969 that have not been read. 8 OCTOBER — THREE OF THE FOUR WOMEN HAVE A CANDIDATE AND A REGISTER NUMBER. The rest of West Berlin was read: Charlottenburg, Steglitz, Tempelhof, Tiergarten, Wedding, Wilmersdorf and Zehlendorf, 1967 to 1994, 193 volumes, no download errors. ELLA — Sterberegister TIERGARTEN 1974 Nr. 1160, “Leibholz geb. Baesler, Ella Josephine Martha”, 22 June 1974. She is in the directory to 1970 and her own borough was read out for 1970-1994 and is empty; the gap is FOUR YEARS. CHARLOTTE — Sterberegister TEMPELHOF 1982 Nr. 161, “Leibholz geb. Schötzow, Charlotte Herta Marie”, 14 January 1982. HILDE — Sterberegister WILMERSDORF 1983 Nr. 304, “Leibholz, Hilde Henriette”, 7 February 1983, AND SHE DIED IN HER OWN BOROUGH: Wiesbadener Straße 57 is Wilmersdorf, she is in nine consecutive telephone books to 1980-81, and the gap is about two years. She is the strongest of the three. AGNES IS THE ONE THIS SWEEP DID NOT ANSWER and Kreuzberg is exhausted 1980-1994, so more volumes is not obviously the next move. ALL THREE CANDIDATES ARE JOINS ON SURNAME, FORENAME AND A BOUNDED GAP — the material this archive got the Bloom mariner wrong on — and each needs ONE CERTIFICATE from a DIFFERENT OFFICE to confirm or kill. THOSE ARE PURCHASES AND DAVID'S TO MAKE. The sweep also produced HERTHA LEIBHOLZ gesch. Katznol (Zehlendorf 1983 Nr. 2733), GERDA FELICITAS BADE geb. LEIBHOLZ (Wilmersdorf 1993 Nr. 312) and MARGARETHE GERTRUD LEIPHOLZ gesch. Driesen (Tempelhof 1967 Nr. 2201). THREE OF THOSE CARRY THE NAME IN THEIR OWN RIGHT, so they had Leibholz PARENTS in Berlin, which is a better kind of lead than a death entry. In data/west-berlin-deaths-2026-10-08.tsv. CLOSED 8 OCTOBER FOR THREE OF THE FOUR. *** ELLA — Tiergarten 1974 Nr. 1160. CHARLOTTE — Tempelhof 1982 Nr. 161. HILDE — Wilmersdorf 1983 Nr. 304, in her own borough. All three are candidates on surname, forename and a bounded gap, each needing one certificate from a different office. AGNES is not found and is carried by row 253. The two post-war Luwinski men are out of reach: the indexes stop at 1994 and the tree has them dying in 1995 and 2006.

DONE 22 September 2026, and not by opening forty-six pages. The search form has an EXPORT AS CSV control; the same request was made inside the page and the response decoded in memory — it is windows-1252, and read as UTF-8 every umlaut is destroyed. It returns EIGHTEEN COLUMNS for all fifty at once: birth name, Geburtskreis, Geburtsland, borough, Inhaftierung, Emigration, both deportation legs with dates, Schicksal, place and date of death. leibholz-gedenkbuch.tsv held four and now holds FIFTY-TWO. THE ROW'S OWN PREMISE NEEDS CORRECTING AND THAT IS THE MOST USEFUL THING HERE: forty-four of the fifty do live in Berlin, but the Gedenkbuch gives the BOROUGH, NOT THE STREET. A borough cannot be laid against the Adressbuch the way the Weinstraße household was; it can only say which pages are worth reading. That work is still to do and is bigger than this row supposed — see row 260. WHY FIFTY-TWO AND NOT FIFTY: a ten-spelling sweep of the same box found LAIBHOLZ, which returns GEORG LAIBHOLZ b. 01.01.1883 and LIDIA LAIBHOLZ b. 1884, both of Berlin, both dead at Auschwitz, and neither a duplicate of the Georg and the Lydia already in the fifty. One vowel had hidden two murdered people from every search run here. AND THE SWEEP CAUGHT A FAULT IN ITSELF: pressing Return in that box does not submit the form, and five terms read back '50 Results' that were all the previous page. Five untested negatives nearly entered this archive as five findings. See gedenkbuch-leibholz-findings.tsv and the negatives page.

THIRTY-TWO bounded negatives are published on this site and most were searched under LEIBHOLZ alone. Three separate institutions have since been shown to write the same murdered person both LEIBHOLZ and LEIPHOLZ. EVERY ONE OF THE THIRTY-TWO IS THEREFORE PROVISIONAL, and the archive has not said so on the page. Two things are owed: first, mark which of the thirty-two could be reached by the eighth spelling and which genuinely could not; second, re-run the ones that could. THIS IS THE SAME FAILURE AS THE UMLAUT AND THE BLANK GENETEKA PARAMETER — a negative is only as wide as the spelling it was searched under, and this archive learned that lesson after it had already printed thirty-two of them. AND IT NOW HAS A WORKED EXAMPLE RATHER THAN AN ARGUMENT. On 21 September the Schöneberg death indexes were scanned for LEIBHOLZ across 1970-1977 and returned nothing four years running. The pattern was widened to Le[iy][bp]h and the 1975 volume answered immediately with ERNA FRIDA ELSE LEIPHOLZ. THE NARROW SCAN HAD ALREADY READ THAT VOLUME AND CALLED IT EMPTY. That is this row's claim demonstrated on a document, not inferred. 21 SEPTEMBER — CLASSIFIED, AND THE FIRST THING IT FOUND WAS AN ERROR IN THIS ROW. This row says 'most were searched under LEIBHOLZ alone'. THEY WERE NOT. Of the thirty-two, ONLY THREE MENTION A LEIBHOLZ SPELLING AT ALL — no. 18 (NAAIRS), no. 28 (the Koszalin letter about Hammerstein) and no. 31 (TNA Discovery). The other twenty-nine are WEAR, LUWINSKI, CASTLE, FIGUEIREDO, SALKOW and record-class negatives that name no person. The row's principle is right and its arithmetic was not, and the corrected shape is more useful: THE QUESTION IS NOT 'WHICH WERE SEARCHED UNDER ONE LEIBHOLZ SPELLING' BUT 'WHICH WERE SEARCHED UNDER ONE SPELLING OF ANY NAME'. On a crude count eighteen of the thirty-two name a single form of their family name, though that over-counts, because several are not name searches at all — no. 12 is about whether a grave reopening was recorded and no. 13 about what a register asked for. WHAT WAS ACTUALLY RE-RUN, AND IT IS ONE. Only no. 31 can be re-run free and without an account today: NAAIRS has stopped answering altogether (row 178), FamilySearch needs a sign-in, and no. 28 is a letter awaiting a reply. TNA DISCOVERY WAS RE-RUN ACROSS FOURTEEN SPELLINGS AND THE NEGATIVE HOLDS AND IS NOW WIDER: LUWENSKI and LUWINSK also return zero, so four forms are bracketed instead of two, and LEIPHOLZ, LEYBHOLZ, LAIBHOLZ, LEIBHOLTZ, LEIPHOLTZ, LEIBOLZ and LEIPHOLS all return zero, so the sixteen Leibholz records are the whole of what Discovery holds under any spelling. Two neighbouring surnames DO return and are now named on the page so nobody finds them twice: LOWINSKI 20, LEWINSKI 228, neither this family. WHAT IS STILL OWED. The eighteen single-spelling negatives need reading one at a time to separate the name searches from the record-class ones, and the name searches re-run when their instrument is reachable again. THE ROW STAYS OPEN and it now says what it is asking for. 22 SEPTEMBER — ALL THIRTY-THREE CLASSIFIED ONE AT A TIME, AND THE ROW'S OWN COUNT COMES DOWN AGAIN. This row had already corrected itself once, from 'most were searched under LEIBHOLZ alone' to a crude count of eighteen single-spelling negatives. EIGHTEEN IS STILL TOO HIGH AND THE REAL NUMBER IS TEN. Reading each one: NINE ARE NOT NAME SEARCHES AT ALL (a coverage count, a structural argument about consular registers, a keyword search on NATURALIZATION, two questions about what a register FORM asked for, a holdings question about Krmpote, a question about what is online, a DNA argument, and an archive's signed answer about whether two series survive) — spelling cannot be the reason for any of those zeros. THREE ARE WITHDRAWN and are not live negatives. ELEVEN WERE ALREADY MULTI-SPELLING. That leaves TEN genuine one-spelling searches for a person. THREE WERE RE-RUN AND THE SCORE IS ONE HELD, ONE DOWNGRADED, ONE CLOSED WITHOUT SEARCHING. No. 3, Leonard's death in England and Wales, was re-run with FreeBMD's own PHONETIC flag: soundex WEAR returns 6,664 deaths for 1970-1985 against 241 for the exact string, TWENTY-SEVEN TIMES the pool the original saw. Narrowed by the forename it gives about forty Leonards across WEIR, WARE, WARR, WRAY, WHEAR, WYER, WERE, WROE, WERRY, WHURR, WIER and WERRA — and NOT ONE WAS BORN IN 1917, which is Leonard's year. THE NEGATIVE SURVIVES AND IS NOW MUCH STRONGER. No. 9, Kurt's naturalisation, needed no search: negative 0 already sweeps the SAME CORPUS under eight spellings with no year limit, so one negative widens the other, and the live re-count still returns exactly eight Luwinski documents. A NEGATIVE CAN BE WIDENED BY READING THE ONE NEXT TO IT. AND NO. 7 WAS DOWNGRADED, WHICH IS THE FIND. The 1973 probate calendar negative was a KEYWORD match over the OCR of scanned calendar pages. Re-run without the keyword, the image half returns FORTY-ONE DOCUMENT PAGES carrying the surname in 1973 and NOT ONE HAS BEEN OPENED; CORRECTED WITHIN THE HOUR, AND THE CORRECTION IS THE REAL FINDING. The first write-up said a Zqxvwyrk null control returned zero and therefore the forty-one were a genuine surname result. IT DOES NOT. Run properly, “41 document pages” comes back for Wear, for Smith, for Luwinski, for nonsense, for 1979 as well as 1973, for the forename Leonard and for the keyword AUBREY. THE IMAGE HALF RUNS NO SEARCH AT ALL — it is a browse handle onto calendar images that opens at BAINBRIDGE. And the digital half holds nothing before 1996 and says so. SO NEGATIVE 7 IS WITHDRAWN, NOT DOWNGRADED: its zero was guaranteed before anyone typed a name. The bad control was published because it agreed with what this archive already suspected and was never repeated. SEVEN REMAIN BLOCKED behind NAAIRS, FamilySearch and Findmypast and are now marked provisional instead of being left to look settled. THE FIRST TO RE-RUN THE DAY NAAIRS ANSWERS IS NO. 18, the Leibholz search, because the eighth spelling is precisely the failure it is exposed to. The whole classification is in negatives-classified.tsv. THE ROW STAYS OPEN on the seven blocked negatives and on the forty-one unread probate pages, which are row 261. AND THE SCORE CHANGES WITH IT: one held, one closed by cross-reference, ONE WITHDRAWN — plus one error of this archive's own, made and corrected the same evening. AND THEN A THIRD TIME, WHICH IS THE ONE THAT IS ACTUALLY RIGHT. The withdrawal above was ALSO WRONG. Find a Will's image half is NOT a browse handle: it calls an OCR endpoint, pp_cowindex_default_search, and called directly it plainly works — WEAR 1973 returns 41 calendar chunks, SMITH 1973 returns 469, LUWINSKI and the nonsense string return 0. WHAT IS BROKEN IS THE PAGE, NOT THE SEARCH: it prints “41 document pages returned” for every one of those, after a fresh load, including with the keyword AUBREY. So negative 7 is NOT withdrawn and NOT proven — IT IS UNRESOLVED, and the two chunks that bracket the name (009223_weake_1973.pdf and 009225_wearmouth_1973.pdf) have never been opened. THREE WRONG ANSWERS IN ONE EVENING AND ONE FAULT BEHIND ALL THREE: reading the rendered page instead of the request underneath it. The first trusted a structurally empty half, the second trusted an unrepeated control, the third trusted a printed number. Each dissolves the moment the endpoint is called and the counts move with the surname. THE RULE: where a page and its own API disagree, the API is the instrument and the page is a report about it. 10 OCTOBER — THE CLASSIFICATION HALF IS DONE, AND THE ROW'S NUMBER WAS WRONG A THIRD TIME. THERE ARE THIRTY-FIVE PUBLISHED NEGATIVES, NOT THIRTY-TWO. Measured mechanically against what each row's own searched field says it tried: EIGHTEEN NAME TWO OR MORE SPELLINGS and stand as published (the best are #31, fifteen forms at TNA Discovery, and #32, ten at the Gedenkbuch); THIRTEEN NAME ONE OR NONE and are provisional; FOUR are not name searches and width does not apply. THE ROW GUESSED EIGHTEEN WERE NARROW. IT IS EIGHTEEN WIDE AND THIRTEEN NARROW — the opposite way round, and the archive is in better shape than the row feared. AND ONE OF THE THIRTEEN IS THE EXACT SHAPE OF THE 1975 ERROR: #18, “no Leibholz appears anywhere in the South African archives”, searched under LEIBHOLZ ALONE — while LEIPHOLZ has produced four people in the Berlin registers this month and has never been put to a South African archive. THE RE-RUNNING HALF IS BLOCKED, NOT UNDONE: #18, #9 and #10 all go through NAAIRS, which row 275 records as blocked. In data/negatives-classified-2026-10-10.tsv. 11 OCTOBER — NAAIRS IS BACK, AND THE THREE NEGATIVES THIS ROW NAMED ARE RE-RUN AND HOLD. The old address is still dead (503 over http, timeout over https, naairs.htm still 404) — THE SERVICE CAME BACK SOMEWHERE ELSE UNDER A DIFFERENT NAME: https://www.nationalarchives.gov.za/search/, the “NATIONAL ARCHIVAL INFORMATION SEARCH SYSTEM”, an Angular portal titled NarsaSearchPortal and subtitled with the very words NAAIRS stands for. Free, no account, ALL TWENTY REPOSITORIES including RSA for all at once. CONTROL FIRST: SMITH 47,936, SCHMIDT 3,418. Then: LEIBHOLZ 0 · LEIPHOLZ 0 · LEIBHOLTZ 0 · LEIPHOLTZ 0 · LAIBHOLZ 0 · LEIBHOLS 0. #18, “no Leibholz appears anywhere in the South African archives”, was the one this row called the exact shape of the 1975 error. IT IS NOW WIDENED TO SIX SPELLINGS AND IT SURVIVES, and the eighth spelling has at last been put to a South African archive. AND #9 AND #10 HOLD TOO: LUWINSKI 7, LUWINSKY 1, LEWINSKI 5, LOWINSKY 7, and seven further spellings zero — with NO NATURALISATION AND NO IMMIGRATION FILE UNDER ANY OF THEM. LUWINSKY returns exactly the one 1948 divorce file the archive has had since September. ALL THREE SHOULD BE RECLASSIFIED FROM PROVISIONAL TO WIDE. AND THE SWEEP FOUND TWO COURT FILES THE ARCHIVE DID NOT HAVE: *** TAB WLD 1878/1956, “RESTITUTION OF CONJUGAL RIGHTS. KURT LUWINSKI VERSUS FRANCES (formerly Sefton, formerly Berner, born HOSTOWSKY)” — A SEVENTH CASE where litigation.json held six, and KURT MARRIED HER ON 29 FEBRUARY 1956 AND SUED HER THE SAME YEAR; and TAB WLD 126/1959, Frances Eisenhardt v Rolf, judicial separation, within a year of her 23 May 1958 remarriage. Both were found under HOSTOWSKY and HOSTOWSKI — ONE LETTER APART AND NEITHER RETURNED BY THE OTHER. Plus the archival shelf mark for that 1958 marriage, KAB HAWC_1/3/58/1/53_183. THREE FAMILY SURNAMES WERE ALSO PUT TO A SOUTH AFRICAN ARCHIVE FOR THE FIRST TIME: KERGER 3 (two of them IMMIGRATION PAPERS), SCHMUL 5, POLLNOW 1 — none connected to anyone here, all recorded as tried and empty. In data/naairs-is-back-2026-10-11.tsv.

DONE 21 SEPTEMBER for the pins and the distances; the dates are carved off to row 259. Sixteen pins were checked against OpenStreetMap Nominatim at one request a second, great-circle distance from the published coordinate, anything over three kilometres read by hand. TWO WERE WRONG. PORTO was published at 41.35 N and Porto is at 41.15 N — ONE DIGIT, TWENTY-TWO KILOMETRES, and it is where Derrick's mother is buried; the move 'Johannesburg to Porto' carried the same wrong figure. SOMMERFELD was 52.95, 12.98 and belongs at 52.80, 13.03, seventeen kilometres, settled from the archive's own tree which gives Sommerfeld/Osthavelland, postcode 16766, Hohenbrucher Chaussee — that is Kremmen. It is where Siegfried was hidden for twenty-seven months. Both fixed. AND THE TRAP, WHICH IS THE MORE USEFUL HALF. The first pass also flagged CHEŁMNO at 4.7 km and AUSCHWITZ at 4.3 km AND BOTH PINS ARE CORRECT. A gazetteer lookup of the modern town is not a check on a historical site: 'Oświęcim' returns the town centre four kilometres from the camp, and 'Chełmno nad Nerem' returns a different Chełmno altogether. Querying the Kulmhof camp and the Auschwitz-Birkenau museum puts both pins within 0.4 km. AN AUTOMATED PASS RUN WITHOUT THAT CORRECTION WOULD HAVE MOVED JACOB, SAMUEL AND OSKAR'S PLACE OF DEATH FOUR KILOMETRES AND GEORG AND LYDIA'S FIVE, in the name of accuracy. A flagged distance is a question, not a finding. THE STATED DISTANCES. There were three on the whole site. 'Some 45 km apart along the Dão' was right for one pair and understated the spread of the three places by nine kilometres — measured, Treixedo to Tondela is 9.4, Tondela to Castro Daire 44.6, Treixedo to Castro Daire 54.0 — and the river was wrong as well, so the sentence now gives the three measured figures and names no river. 'Ivone — 400 miles of railway' is LEFT ALONE AND MARKED UNVERIFIED: the straight line is 282 miles, the claim is about a rail route that detoured north through Komatipoort and Waterval Boven, and replacing one unsourced number with another is not an improvement. The third, 'about fifteen kilometres' for Bischofstein to Sauerbaum, is the error that started this row and was already corrected to 17.8. Full working in data/atlas-pin-audit.tsv.

The name index gives the date (13 January 1975), the register number (79) and the confession (ev). A DEATH ENTRY OF THAT VINTAGE CARRIES THE ADDRESS, THE BIRTH DATE OR AGE, THE BIRTHPLACE, THE MARITAL STATUS AND THE INFORMANT -- and the informant is usually the nearest relative, which is the thing this archive keeps running out of. THE TENSION TO RESOLVE FIRST, AND IT COSTS NOTHING: this archive records "Leibholz Erna · 20 Streitstr. 39" in the 1980-81 telephone book as the last Leibholz in the post-war run. A woman who died in January 1975 is not in a 1980-81 telephone book. EITHER THEY ARE TWO WOMEN OR A YEAR HAS BEEN MISREAD, and re-reading the Streitstraße line settles it before anybody orders a certificate. The certificate itself is a purchase and is not mine. 21 SEPTEMBER, RESOLVED, AND THE TIDY ANSWER WAS WRONG. The Streitstraße line was re-read as this row asked. IT IS NOT A LEIBHOLZ AT ALL -- the 1980-81 telephone book OCR read a p as a b, and the 1975-76 and 1979-80 volumes print LEIPHOLZ ERNA plainly, same address, SAME TELEPHONE NUMBER 3 35 84 05 in all three years. See correction 9500904. AND THE DATES DO NOT COLLAPSE INTO ONE WOMAN: Erna Frida Else Leipholz died 13 January 1975 and Erna Leipholz of Streitstraße is listed on her own line for six years after that. TWO WOMEN OF THE SAME NAME. The certificate is no longer needed to tell them apart, which is the cheap half of this row done without spending the expensive half.

PAUL, Schlosser, S59 Dieffenbachstraße 58, from the 1924 Adressbuch. KLARA, PUTZMACHERIN, SW 29 Dieffenbachstraße 58, from the 1957 and 1959 post-war Namenteil. SAME STREET, SAME HOUSE NUMBER, THIRTY-THREE YEARS AND A WAR APART; the postal prefix differs because Berlin renumbered its districts, not because the address did. A Putzmacherin is a MILLINER and the Leibholz trade in Berlin was hats. TESTS THAT NEED NOBODY'S HELP: the intervening volumes 1925-1943 and 1949-1956 for the same house, which would show whether the family held it straight through; the Kreuzberg registry indexes for both names; and the 1943 and 1949 volumes either side of the gap, which is where a household either survives or does not. NOT KIN ON ANY EVIDENCE HELD. CONFIRMED AND JOINED TO ROW 251, 22 SEPTEMBER. The 1957 and 1959 directories print the two together, Friedrich first and then the repeat-dash for the same surname: “Leipholz Friedrich Angest Tgl Berlin — KLARA PUTZMACHERIN SW 29 DIEFFENBACH[STRASSE]”. The adjacency is only alphabetical and proves no kinship, BUT IT SHOWS THESE TWO ARE THE WHOLE OF THE LEIPHOLZ NAME IN BERLIN IN THOSE YEARS. The 1924 Paul at the same house was not reached by this instrument: the only “Leipholz Paul” hits are the directory's own cross-reference line, “Leibholz, s. a. Leipholz”. 10 OCTOBER — THE GAP IS BRIDGED. PAUL IS AT DIEFFENBACHSTRASSE 58a IN 1941, 1942 AND 1943. The 1941 Namenteil line was READ AT THE IMAGE in Fraktur — “Paul Schlosser SW 29 Dieffenbachstr 58a” — and the 1943 TEIL III street page lists house 58a with its owner, its manager and twenty-eight tenants, one of them “LEIPHOLZ, P., SCHLOSSER”. With Paul at 58 in 1924 and Klara at 58 in 1957 and 1959, THE HOUSEHOLD HOLDS THE HOUSE FROM 1924 TO 1959, thirty-five years across the war, and the row's S 59 / SW 29 renumbering reasoning is now witnessed rather than argued. TWO METHOD FAULTS CAUGHT: a full-text search for Leipholz + Dieffenbach returns 1924, 1928 and 1942 AND NOT 1941 — the 1941 OCR mangles the street to “Diessenbachsir”, and the year was found only by reading the run headword-to-next-surname. And THE HOUSE NUMBER IS WRONG IN THE TEXT LAYER FOR EVERY YEAR: the OCR reads 68a, the image says 58a, a Fraktur 5 read as a 6 — the same class of error as the Fraktur 4 read as a 1 in Kreuzberg two days ago. KLARA HAS NO WARTIME ENTRY at all, checked 1939-1943 year by year, zero every time. Still not kin to anyone here. In data/dieffenbachstrasse-58a.tsv. 10 OCTOBER — KREUZBERG 1938-1979 SWEPT, AND KLARA'S WINDOW IS NOW SEVENTEEN TWENTY-FIRSTS CLOSED. Forty-five volumes, every page, no errors: SEVENTEEN LEGIBLE AND TWENTY-EIGHT BLIND, and THE TEXT LAYER BEGINS AT 1963. The prediction going in was that all forty-five would be blind — reasoned from 1983 being blind and 1984 legible — AND IT WAS WRONG BY SEVENTEEN VOLUMES. No Leibholz and no Leipholz in any legible year 1963-1979. Two LOWINSKI women came out, both almost certainly the Polish Catholic strand: Luise Emma verw. Furtmann (1969 Nr. 277, kath.) and Ingrid geb. Palm (1977 Nr. 324). AND 1959 IS READ AT THE IMAGE: L block pages 144-156, bounded by K on 143 and M on 157, Januar to Dezember, NO LEIPHOLZ — near misses Liepold verehl. Morgenthal, Lippold geb. Hansen and Leipolt verw. Fendor. KLARA'S WINDOW IS 1959-1979 AND SEVENTEEN OF ITS TWENTY-ONE YEARS ARE NOW CLEAR. WHAT REMAINS IS EXACTLY THREE BLIND VOLUMES — Kreuzberg 1960, 1961, 1962, Nr. 679, 680, 681 — and the L block sits at about 45 per cent through, so the next session can go straight to it. In data/kreuzberg-1938-1979-deaths.tsv. 11 OCTOBER — KLARA IS FOUND, IN THE FIRST OF THE THREE VOLUMES THAT WERE LEFT. “LEIPSHOLZ geb. DIEDRING · JOHANNE KLARA · 4. · – · 316” — STERBEREGISTER KREUZBERG 1960 NR. 316, died 4 FEBRUARY 1960, no confession stated. P Rep. 510 Nr. 679 p.123 under the Februar heading, read at the image, with her neighbours dating the month beyond doubt. AND THE SURNAME IS LEIPSHOLZ, WITH AN S — A NINTH SPELLING, read at five times magnification and a registrar's own hand, not a machine's. STRONG CANDIDATE FOR THE KLARA LEIPHOLZ OF DIEFFENBACHSTRASSE 58: the forename is Klara, the borough is Kreuzberg, and she died in THE FIRST BLIND YEAR AFTER HER LAST DIRECTORY ENTRY — 1959 read at the image and empty, 1963-1994 swept and empty. NOT CLAIMED: the spelling differs by a letter and “geb. Diedring” is a maiden name never held here. KREUZBERG 1960 NR. 316 GIVES THE LAST ADDRESS; if it reads Dieffenbachstraße 58 the question closes and with it the household's thirty-five-year run, 1924 to 1960. She is on the register as the eighth documented person. 1961 and 1962 remain unread and are now MOOT FOR HER — worth reading only to complete the office. In data/klara-leipsholz-kreuzberg-1960.tsv. 11 OCTOBER, LATER — 1961 AND 1962 ARE READ, AND KREUZBERG IS CLOSED. Both are blind volumes and both were read page by page at the image: 1961 L block PAGES 118-128, 1962 L block PAGES 118-129, each running Januar to Dezember and each bounded by K closing before and M opening after. NO LEIBHOLZ, NO LEIPHOLZ, NO LEIPSHOLZ IN EITHER. Near misses settled by eye: Lippoldt verw. Schuboth, Lippke verehl. Schenck, Lippert geb. Schmidt, Lipowsky verw. Wagner (1961); Liebe geb. Lehmann, Liepe geb. Siegfried, Lipp geb. Hahn, Lipowy geb. Wiese, Leupold, Leopold geb. Spanowski, Lichtnack geb. Bartsch (1962). THEY PROVE NOTHING ABOUT KLARA — she was already dead when 1961 opened — AND THAT IS THE POINT OF SAYING SO. They were read so that the sentence “Kreuzberg deaths 1938 to 1994 hold exactly two of our people” is a measured statement about fifty-seven years and not a sufficient one about forty-nine. THE OFFICE IS COMPLETE. WHAT IS LEFT ON THIS ROW IS A PURCHASE AND IT IS NOT MINE TO MAKE: Kreuzberg 1960 Nr. 316 gives Klara's last address. If it reads Dieffenbachstraße 58 the household's thirty-five-year run closes, 1924 to 1960, and a ninth spelling of the surname is tied to a street this archive has been working for three weeks. If it reads anything else, she is a different Klara and the row ends honestly instead of hopefully.

Tegel, Berliner Straße 5a, Angestellter, 1957 and 1959. Reinickendorf, Schwabstraße 4, PENSIONÄR, 1961 through 1970, with the same neighbours in the street section both times. A man who retires between 1959 and 1961 was born around 1895 give or take, WHICH PUTS HIM IN JACOB LEIBHOLZ'S GENERATION. Reinickendorf's death indexes are in the same free LABSA set that gave up Erna, and his death would name a wife and an age. CHEAP, UNASSISTED, AND NOT YET DONE. 21 SEPTEMBER: REINICKENDORF 1974-1981 READ, EIGHT CONSECUTIVE YEARS, HE IS IN NONE OF THEM. All eight bracketed -- the nearest name before and after the gap is recorded for every year in data/kreuzberg-reinickendorf-deaths.tsv. 1982-1994 still to read, and those volumes jump from 3 MB to 99 MB because the series changes format. EXTENDED AT BOTH ENDS 22 SEPTEMBER. The run is not 1961-1970, it is 1957 TO 1973-74. He is an ANGESTELLTER at Tegel in 1957 and 1959, a PENSIONÄR from 1961, at Schwabstraße 4 from at least 1967 — and he is in the TELEPHONE BOOKS at that flat in 1970-71 (49 80 11) and again in 1973-74 (4 91 82 51). TWO NUMBERS AT ONE ADDRESS: Berlin was lengthening subscriber numbers across that period, which is the ordinary explanation, but 49 80 11 does not expand tidily into 4 91 82 51 and no claim is made. 22 SEPTEMBER — BRACKETED, AND THE NAME SEARCH THAT WAS USED TO DO IT WAS WRONG. All twelve telephone volumes 1969-70 to 1980-81 were read AT THE ALPHABETICAL SLOT between LEIPERT and LEIPIC, which is the row 252 method transferred to these books, and it had to be: A FULL-TEXT SEARCH FOR THIS SURNAME IS NOT AN INSTRUMENT. The OCR prints it LEIPHOTZ in 1972-73, LELPHOLZ in 1976-77 and 1978-79, LEIPKSLZ in 1977-78 and LEIBHOLZ in 1980-81 — FIVE OF THE ELEVEN READABLE VOLUMES MANGLE IT, and a search for the right spelling misses every one. A healthy SCHMIDT control (133-191 a volume) sat over the top of that the whole time: A COMMON-NAME CONTROL CANNOT CLEAR A RARE-NAME NEGATIVE. THE RESULT: HE IS IN 1970-71, 1972-73 (A VOLUME THIS ARCHIVE DID NOT HAVE) AND 1973-74, AT SCHWABSTRASSE 4 IN ALL THREE, AND THE SLOT IS COMPLETE AND EMPTY IN 1974-75. His listing ends between the 1973-74 and 1974-75 books — a year tighter than before. 1971-72 has no text layer at all and is inconclusive, not a gap. AND THE BRACKET DOES NOT AGREE WITH THE REGISTERS. He leaves the book during 1974 and REINICKENDORF 1974-1981 IS ALREADY READ IN FULL AND HE IS IN NONE OF IT. So he moved and died elsewhere, or he died in Reinickendorf after 1981, or he gave up the telephone and lived on — and Heinrich Leibholz, who vanished from the directory in 1942 and did not die until 1948, forbids the easy reading. NEXT: Reinickendorf 1982-1994, the 99 MB volumes. Full working in data/leipholz-telephone-slot-audit.tsv. ERNA OF STREITSTRASSE 39 came out of the same sweep CONTINUOUS ACROSS ALL SIX VOLUMES 1975-76 TO 1980-81 on one number, where the archive had three — which hardens the two-Ernas finding rather than overturning it. 23 SEPTEMBER — THE REMAINING WORK IS THIRTEEN VOLUMES. The full LABSA catalogue puts Reinickendorf's death indexes for 1982-1994 at THIRTEEN, not the open-ended run this row describes. And if he moved, the whole of 'somewhere else in Berlin' is 296 death-index volumes across 25 offices — bounded, and listed in data/labsa-catalogue.json. 23 SEPTEMBER — THE THIRTEEN REMAINING VOLUMES ARE READ. REINICKENDORF IS FINISHED AND HE IS NOT IN IT. P Rep. 130 Nr. 1478-1490, 1982 to 1994, read through pdf.js over HTTP range requests — 807 pages of 2,120, nothing downloaded. ALL THIRTEEN BRACKETED ON BOTH SIDES OF BOTH SPELLINGS. That mattered: THE FIRST PASS BRACKETED ONLY THE LEIBHOLZ SLOT and would have published thirteen negatives covering LEIPHOLZ too on a test that never looked at the Leipholz slot — in 1990 the nearest name above Leibholz is LEIPE, which sorts BELOW Leipholz. The pass was discarded and the volumes read again. Two volumes, 1992 and 1993, failed mid-read with network errors on the first pass and were NOT counted as absences; both read clean on the retry. SO TWENTY-ONE CONSECUTIVE YEARS OF HIS OWN BOROUGH, 1974 TO 1994, ARE NOW READ AND HIS DEATH IS NOT REGISTERED THERE. He either moved — 296 death-index volumes across 25 offices, bounded now — or died after 1994 at nearly a hundred. AND THE SWEEP FOUND A PERSON NOBODY HERE HAD: LEIPHOLZ MARIE, Reinickendorf Sterberegister 1988 Nr. 2306, the ONLY Leipholz in those twenty-one years, dying in the borough Friedrich lived in. Not either Erna. No Marie Leipholz is held in this archive under any spelling. READ AT THE IMAGE THE SAME DAY, AND IT CORRECTED THE TEXT LAYER. The row is typewritten and plain: "Leipholz | Marie | 05.11. | 2306 | ev." — crop at /leipholz-marie-1988.jpg. THE DATE IS 5 NOVEMBER, not the 5 July the OCR's "@507" suggested; the guess was wrong by four months and was published as a guess rather than a date, which is the only reason it did no harm. AND THE IMAGE GAVE A COLUMN THE LAYER NEVER SHOWED: SHE IS EV., EVANGELISCH. The same column reads mosaisch for the Jewish woman in the Spandau 1982 index, and reads mosaisch for LEINKRAM GERTRUD three lines above Marie on her own page — so the distinction is live on that very sheet. A PROTESTANT LEIPHOLZ SITS WITH THE MASURIAN LEIPHOLZ FAMILY OF PIANKI AND ARYS, whom this archive already holds and already says are certainly not the same people, and not with Jacob Leibholz of Schubin. One word, no proof, and it points away rather than toward. Her birthplace and informant still need the certificate, WHICH IS A PURCHASE AND NOT THIS ARCHIVE'S TO MAKE. Full working in data/reinickendorf-1982-1994-deaths.tsv. 7 OCTOBER — REINICKENDORF IS NOW READ WITHOUT A GAP FROM 1974 TO 1994, TWENTY-ONE CONSECUTIVE YEARS, AND HE IS IN NONE OF IT. 1982-1994 read 7 October, every page, all thirteen volumes legible (0.93 to 1.75 control names a page) and bracketed. The borough he lived in has now been asked the whole question and has answered no. BUT THE RUN GAVE UP A LEIPHOLZ AND IT IS NOT HIM: MARIE LEIPHOLZ, Sterberegister Reinickendorf 1988 Nr. 2306, ev. AND SPANDAU, THE NEXT BOROUGH WEST, GAVE UP A FRITZ. FRITZ LEIPHOLZ, Sterberegister Spandau 1985 Nr. 2926, ev., VERWITWET. Fritz is the ordinary short form of Friedrich; Friedrich leaves the Berlin telephone book during 1974; Reinickendorf does not hold him. THAT IS A SHAPE AND NOT A DOCUMENT — no birth date, no address, no wife's name — and it is written down as a candidate precisely because it fits too well to be trusted on its own. THE CERTIFICATE WOULD SETTLE IT and it is orderable by number. Note also that Marie is NOT his widow: he is entered verwitwet in 1985 and she died in 1988. In data/other-offices-deaths-2026-10-07.tsv. CLOSED 8 OCTOBER. Reinickendorf is read without a gap 1974-1994, twenty-one years, and FRIEDRICH IS IN NONE OF IT. The run gave MARIE LEIPHOLZ (Reinickendorf 1988 Nr. 2306) who is not him, and Spandau gave FRITZ LEIPHOLZ (1985 Nr. 2926, ev., VERWITWET) who may be — Fritz is the ordinary short form of Friedrich and Spandau is the next borough west. A candidate, not an identification; the certificate settles it.

Seven of sixty-one are read. The method is now routine and costs about fifteen seconds a volume: LABSA lists the volumes, pdf.js reads them in place over HTTP range requests without downloading a byte of the 40-380 MB files, and the test is not a name search but PULLING EVERY Le[hij]* TOKEN OUT OF THE MIDDLE THIRD AND READING THE ALPHABET FOR A GAP. THE PATTERN MUST BE Le[iy][bp]h AND NOT LEIBHOLZ -- that is the whole lesson of 1975. Then Kreuzberg for Agnes and for Dieffenbachstraße, Neukölln for Charlotte, Spandau for Erna, Reinickendorf for Friedrich, Wedding for Joachim the Ofensetzer. AND THE METHOD NEEDED A GUARD BEFORE IT COULD BE TRUSTED AT SCALE. Carrying the Schöneberg band (33-64 per cent) over to Kreuzberg produced TWO FALSE NEGATIVES -- 1980 stopped scanning at Leipelt and 1981 at Leimbach, both of which sort BEFORE Leipholz, so each reported an absence it had never looked for. A scan must now return the nearest name on EACH SIDE of the target, and a result with an empty side is inconclusive rather than negative. It caught a third case within minutes. ALSO: a volume much larger than its neighbours may hold TWO alphabetical sequences -- Kreuzberg 1982 is 226 pages against 125 and has L twice. And a binary search keyed on a page's median word DOES NOT WORK here, because these registers print forenames in their own column. PROGRESS 21 SEPTEMBER: Schöneberg 1970-1994 complete (25 volumes), Reinickendorf 1974-1981 (8), Kreuzberg 1980-1982 (3), Spandau 1982 (1). THIRTY-SEVEN VOLUMES READ IN A DAY, none downloaded. Remaining: Schöneberg 1938-1969, Kreuzberg 1983-1994, Reinickendorf 1982-1994, Spandau 1983-1994 and all of Neukölln and Wedding. CORRECTED 22 SEPTEMBER, AND THE CORRECTION IS THAT THE WORK WAS ALREADY DONE. This row opens “Seven of sixty-one are read” and closes by listing “Schöneberg 1938-1969” as remaining. NEITHER IS TRUE ANY LONGER, AND HAS NOT BEEN SINCE 21 SEPTEMBER. The resume pointer — schoeneberg-progress.tsv, which exists precisely so that a session reads it before starting — states it plainly: “COMPLETE. 1938 TO 1968 IS READ AT THE IMAGE WITHOUT A GAP, AND 1970-1994 BY TEXT LAYER”, thirty-one years page by page, 295 pages of L surnames, roughly 9,900 names. AND 1969, THE ONE YEAR COVERED BY NEITHER METHOD, IS RECORDED SEPARATELY AS “READ IN FULL” on the same date. ALL SIXTY-ONE SCHÖNEBERG VOLUMES ARE DONE. Two hits in the whole run: Günther Leibholz 1939 Nr. 780, and Erna Frida Else Leipholz 1975 Nr. 79. SO THIS ROW IS NOT A SWEEP OF SCHÖNEBERG, IT IS THE OTHER OFFICES, and it has been retitled to say so. WHAT ACTUALLY REMAINS: KREUZBERG 1983-1994 (which is row 253), REINICKENDORF 1982-1994, SPANDAU 1983-1994, and ALL OF NEUKÖLLN AND WEDDING. The method, the Le[iy][bp]h pattern, the two-sided-neighbour guard and the two-alphabets-in-one-volume warning all stand and are the reason the Schöneberg run can be trusted. THE LESSON IS THE ROW'S OWN: a resume pointer only works if the row that sends you to it is updated when it closes, and this one was not. 7 OCTOBER — FIFTY-ONE VOLUMES READ IN ONE SITTING, AND THE ROW IS NOW MOSTLY CLOSED. KREUZBERG 1983-1994, REINICKENDORF 1982-1994, SPANDAU I 1983-1994 AND ALL OF NEUKÖLLN 1967-1992. FOUR PEOPLE OF THE NAME CAME OUT: FRITZ LEIPHOLZ, Spandau 1985 Nr. 2926, ev., VERWITWET — a candidate for Friedrich of Schwabstraße 4 and not an identification · MARIE LEIPHOLZ, Reinickendorf 1988 Nr. 2306, ev. — NOT his widow, because he is already verwitwet in 1985 · GERTRUD ELLA MEIRITZ née LEIPHOLZ, Spandau 1993 Nr. 2047 · and LEOKADIA LOWINSKI née BARANIAK, Neukölln Nr. 815, kath., who is almost certainly the POLISH CATHOLIC strand and is recorded for that reason. THE METHOD CHANGED IN TWO WAYS AND BOTH MATTER. FIRST, EVERY PAGE OF EVERY VOLUME IS READ, not the middle third — which dissolves the band fault that produced the Kreuzberg 1980 and 1981 false negatives and the two-alphabets fault with it. SECOND, AND THIS ROW WOULD HAVE PUBLISHED TWELVE WORTHLESS NEGATIVES WITHOUT IT, EVERY VOLUME NOW SCORES ITS OWN TEXT LAYER against six common surnames before its silence counts: Kreuzberg 1983 scores 0.095 a page and Kreuzberg 1984 scores 1.272, and the first pass of this sweep had returned a clean bracket for 1983 built on a layer row 253 had already proved blind. THREE VOLUMES CAME BACK BLIND — Kreuzberg 1983 (already read at the image), Neukölln Nr. 710 and Nr. 822 (both since read at the image) — AND ONE IS STILL UNREAD: SPANDAU P Rep. 480 Nr. 2458, 1982-1986, which is the only loose end of the fifty-one. AND A THIRD LESSON, FROM NEUKÖLLN 1967-68: THE BRACKET IS NOT A PROOF OF COVERAGE IN EVERY BOOK. That volume runs BY MONTH AND DATE inside each letter, not alphabetically, so nothing sorts either side of anything. Reading the whole letter is the only test that survives. In data/other-offices-deaths-2026-10-07.tsv. 8 OCTOBER — WEST BERLIN IS NOW COMPLETE EXCEPT FOR FIVE BLIND VOLUMES. All twelve boroughs read: the five of 7 October plus Charlottenburg, Steglitz, Tempelhof, Tiergarten, Wedding, Wilmersdorf and Zehlendorf for 1967-1994 — 193 more volumes, 17.8 GB, zero errors. EAST BERLIN IS DELIBERATELY NOT DONE and that is a decision rather than an omission: a West Berliner in a 1979 telephone book did not move to the DDR, and the remaining 467 volumes there would be sweeping to look thorough. THE METHOD TOOK TWO MORE CORRECTIONS. (1) The scanner had NO RETRY on its downloads and lost its opening volume to a dropped connection — which would have made a volume silently missing from a sweep reporting itself complete. Retry, resume and a stall timeout added. (2) TEMPELHOF 1967-68 is the SECOND volume found whose L block runs BY DATE AND NOT ALPHABETICALLY, after Neukölln 1967-68. In a book ordered like that the two-sided bracket proves nothing and only reading every page of the letter works. FIVE VOLUMES CARRY NO TEXT LAYER AND ARE RECORDED AS UNREAD, NOT AS NEGATIVES: Charlottenburg 1990, Tempelhof 1914-1979, Tiergarten 1967, Tiergarten 1968, Wilmersdorf 1990. Tempelhof 1967-68 was blind too and HAS been read at the image — pages 136-149, both year headings read — because it is the one volume whose silence could have undercut the Charlotte candidate. It holds one Leipholz and no Leibholz. 8 OCTOBER, LATER — ALL SIX BLIND VOLUMES READ AT THE IMAGE. WEST BERLIN IS 193 OF 193. TIERGARTEN 1967 (pp. 84-92) and 1968 (pp. 85-93), both NOT alphabetical inside the letter and both read page by page — and they matter because Tiergarten is where ELLA was found in 1974, so a second Leibholz there in 1967 or 1968 would have weakened her. There is none. WILMERSDORF 1990 settled on ONE page: Lehmbeck then Leibelt, Leipacher then Leipold, both slots bracketed tight. CHARLOTTENBURG 1990 (pp. 92-96): Lehnhäuser then Leibner, Leider then Leipziger. TEMPELHOF 1914-1979 (pp. 147-162) is the odd one — a single retrospective volume covering SIXTY-FIVE YEARS in KURRENT, arranged by letter then by year, with a religion column; nothing in it. And TEMPELHOF 1967-68 (pp. 136-149), read earlier, gave MARGARETHE GERTRUD LEIPHOLZ gesch. Driesen, 5 November 1967, Nr. 2201. THE SIX ARE 1,270 PAGES NO TEXT SEARCH CAN EVER TOUCH, in five hands and four arrangements. Left out, the sweep would have been 97 per cent of West Berlin reported as 100 — and one of them held a Leipholz nobody could have found by searching. CLOSED 8 OCTOBER. ALL TWELVE WEST BERLIN BOROUGHS ARE READ, 1967-1994. 244 volumes between the two sweeps, every page of every volume, six blind ones read at the image. Year coverage verified borough by borough: eleven complete, Neukölln missing only 1993-94. East Berlin deliberately not swept and the reason is written down. data/west-berlin-deaths-2026-10-08.tsv.

She is in the telephone book at Mehringdamm 109 from 1972-73 to 1979-80 and gone from 1980-81, which made 1980-82 the obvious years. SHE IS IN NONE OF THEM, all three bracketed. Kreuzberg 1983-1994 remain, and so does the possibility that she did not die in Kreuzberg at all. THE CONTROL CASE THIS ARCHIVE ALREADY OWNS IS THE WARNING: Heinrich Leibholz, her husband, vanished from the Berlin directory after 1942 and DID NOT DIE UNTIL 1948. Leaving a telephone book is not dying. She was born 10 March 1889, so she was 91 in 1980 -- a move into a home in another borough is at least as likely as a death, and a home means another office. 22 SEPTEMBER — 1983 BRACKETED, AND THE METHOD CORRECTED BEFORE IT COULD DO HARM. All twelve Kreuzberg volumes for 1983-1994 are labelled 'Durchsuchbares PDF durch OCR-Texterkennung', which promised the whole row in minutes. THE LABEL IS NOT TRUE OF THE NAMES. A frequency control on Kreuzberg 1983 — 211 pages, a whole year of a borough — finds SCHMIDT 6, MÜLLER 7, LEHMANN 4, LANGE 1 AND BECKER ZERO in the text layer. The registers are handwritten and the OCR has caught the printed headings and almost nothing else. The page image settles it: page 106, the first page of M, carries FIVE MÜLLERS IN JANUARY ALONE, nearly as many as the OCR found in the entire year. A text search here returns a clean nothing and the nothing is the instrument, not the record. AND THE SAME TEST CLEARED TWENTY-FIVE PUBLISHED YEARS. Schöneberg 1970-1994 was closed 'by text layer' and that closure is SOUND: 1975 gives Schmidt 31, Müller 29, Becker 7 and 1988 gives 29, 24, 8 — four to five times Kreuzberg's yield on fewer pages — and the 1975 layer returns exactly one Le[ip]h match, which is Erna Frida Else Leipholz, the find already held. An instrument that finds the one thing that is there is one whose silence means something. SO 1983 WAS READ AT THE IMAGE. L/1983 is pages 96-105, bounded by the tail of K on 95 and the opening of M on 106, all twelve months present and in order, about 150 L surnames. NO LEIBHOLZ, NO LEIPHOLZ, NO KERGER. The nearest are Liebsch, Liebau, Liebert, Liebe and Leyser. The twice-filing of married women runs here too — 'Lachmann geb. Lipinski' and 'Lipinski verh. Lachmann', both Nr. 4/83 — so the L block covers a Leibholz-born woman who married out. 1984-1994 REMAIN AND THE COST IS NOW KNOWN: eleven volumes, about ten L pages each, two legible half-screenshots per page, roughly 220 reads. Bounded and dull and the only honest route. AND THE PRIOR QUESTION IS UNCHANGED — she was 91 and left the telephone book; a home in another borough is at least as likely as a death in Kreuzberg, and reading eleven more Kreuzberg volumes tests one branch of that and not the other. Full working in data/kreuzberg-ocr-control.tsv. 23 SEPTEMBER — 'ANOTHER OFFICE' IS NOW A NUMBER. This row raises the possibility that Agnes, 91 in 1980, moved into a home in another borough and died there. The full LABSA catalogue makes that exactly 296 DEATH-INDEX VOLUMES ACROSS 25 OFFICES that reach 1982 or later; Kreuzberg von Berlin's own share of it is THIRTEEN. Not a vague elsewhere. 7 OCTOBER — CLOSED, AND THE 220 IMAGE READS THIS ROW WAS SAVING UP FOR ARE NOT NEEDED. This row assumed Kreuzberg stayed handwritten to 1994 and planned to read eleven volumes at the image. IT DOES NOT. The series changes at 1984: the volume drops from 125 MB to 33 MB, the register becomes a TYPED INDEX, and the text layer's control score jumps from 0.095 common surnames a page to 1.272 — thirteen times. ALL ELEVEN VOLUMES 1984-1994 WERE READ BY TEXT, EVERY PAGE, EVERY ONE BRACKETED ON BOTH SIDES, AND AGNES IS IN NONE OF THEM. With 1980-1982 (read 21 September) and 1983 (read at the image on 22 September), KREUZBERG IS NOW COMPLETE 1980 TO 1994 AND SHE IS NOT IN IT. WHAT THAT DOES AND DOES NOT SETTLE: she left the telephone book at Mehringdamm 109 after 1979-80 and she was ninety-one. THE BOROUGH IS EXHAUSTED; THE QUESTION IS NOT. A move into a home in another borough is at least as likely as a death, and this archive's own control is Heinrich Leibholz, who vanished from the directory after 1942 and did not die until 1948. The row goes to the homes and the neighbouring offices, not to more Kreuzberg. In data/other-offices-deaths-2026-10-07.tsv. 8 OCTOBER — SHE IS STILL NOT FOUND, AND THE SEARCH IS NOW BOUNDED INSTEAD OF OPEN. SHE DID NOT DIE IN WEST BERLIN BETWEEN 1967 AND 1994. All twelve boroughs are read, 244 volumes, every page, six blind ones at the image — AND THE YEAR COVERAGE WAS VERIFIED RATHER THAN ASSUMED, because a MISSING VOLUME IS INVISIBLE: eleven boroughs are complete year by year and NEUKÖLLN HAS NO INDEX FOR 1993 OR 1994, the only hole in the sweep (she would be 104). AND THE THREE LIKELIEST VOLUMES IN BERLIN WERE RE-SCORED BEFORE BEING TRUSTED: she leaves the telephone book between 1979-80 and 1980-81, so Kreuzberg 1980, 1981 and 1982 are the window — and all three were read in SEPTEMBER, BEFORE the legibility control existed, in the same office whose 1983 volume later proved BLIND. Rescored: 1.717, 1.480 and 0.341 against a floor of 0.15. ALL LEGIBLE, NEGATIVES STAND — though 1982 is the weakest legible volume in either sweep and is flagged as such. WHAT WAS GAINED: A BIRTHPLACE. Mapping the Lives, read at the source instead of from the line this archive had copied: AGNES KERGER WAS BORN 10 MARCH 1889 AT MALLMITZ, KREIS SPROTTAU, SILESIA — today Małomice in Poland, which means she had no home region left to return to after 1945. Her census card is VZ153403, and her “race” is NNNN: FOUR NON-JEWISH GRANDPARENTS. She is in a database of the persecuted ONLY BY MARRIAGE, which is exactly why the instruments that find this family's dead do not reach her. AND ONE NEW NEGATIVE THAT IS A LEAD: HEINRICH IS NOT IN NEUKÖLLN'S 1948 DEATH INDEX, THOUGH HE LIVED THERE. P Rep. 380 Nr. 700, 782 pages, no text layer; L/1948 is pages 357-371, Januar through Dezember with the AUGUST heading on 367, read at the image, and no Leibholz in any of it. A DEATH REGISTERS WHERE IT HAPPENS, NOT WHERE THE DEAD PERSON LIVED — so he died outside Neukölln, most likely in a hospital. HIS ENTRY WOULD NAME THE SURVIVING WIDOW AND HER ADDRESS, which would be the first document about Agnes beyond a census card and a directory line. THE JOB IS BOUNDED: the 1948 index of the other eleven boroughs, one volume each. In data/agnes-leibholz-kerger.tsv. 8 OCTOBER, LATER — HEINRICH IS FOUND, AND HE DIED IN AGNES'S BOROUGH NINE YEARS BEFORE SHE APPEARS IN IT. STERBEREGISTER KREUZBERG 1948 NR. 2493, “Leibholz · Heinrich · 2493 · 9.” under the AUGUST heading of P Rep. 510 Nr. 667, page 180, confirmed at the image. The day column runs in step with the register numbers either side of him, so the bare “9.” is 9 AUGUST 1948 — THE SAME DAY MAPPING THE LIVES GIVES, now said by the register in its own hand, with an office and a number for the first time. HE LIVED IN NEUKÖLLN AND DIED IN KREUZBERG, which is why Neukölln's 1948 index (read page by page the same day, L/1948 pp. 357-371) is correctly empty: a death registers where it happens. The Urban-Krankenhaus stands in Kreuzberg on the Neukölln boundary, which is the ordinary reading and is NOT asserted — the certificate gives the place of death. WHAT IT DOES FOR THIS ROW: Agnes is at MEHRINGDAMM 109, KREUZBERG from 1957 to 1979-80 and this archive never knew why the family was in that borough — Neukölln in 1939, Kreuzberg by 1957, nothing in between. THE CONNECTION BEGINS WITH HIS DEATH. AND A 1948 DEATH ENTRY NAMES THE SURVIVING SPOUSE, THE INFORMANT AND THE ADDRESS, so Kreuzberg 1948 Nr. 2493 would be THE FIRST DOCUMENT ABOUT AGNES BEYOND A CENSUS CARD AND A DIRECTORY LINE. A purchase, and David's. THE SWEEP THAT FOUND HIM: all of Berlin, because the Magistrat did not split until November 1948 — 29 offices, 41 volumes, 6.5 GB, of which 40 ARE BLIND. The scan's value was its map, not its hits: it said which volumes can never be read by text, so the search was RANKED rather than ground through — his own borough's unread second volume, then WEDDING for the Jüdisches Krankenhaus where Gertrud died in 1943, then the adjacent hospital boroughs. He was in the third opened. AND THE LIMIT IS NOW CLOSED: THE WHOLE KREUZBERG 1948 L BLOCK IS READ, pages 170 to 185, bounded by the close of K on 169 and the opening of M on 186, Januar through Dezember — and HEINRICH IS THE ONLY LEIBHOLZ IN THE YEAR. Nearest in the block: LINDHOLZ geb. Thom, one letter away. In data/heinrich-leibholz-kreuzberg-1948.tsv. 11 OCTOBER — THE LAST GAP IS FILLED, AND IT WAS IN HER OWN WINDOW. This archive has said since 8 October that all twelve West Berlin boroughs were swept. THEY WERE — BUT NOT ALL FROM THE SAME YEAR, and nobody had laid the twelve coverages side by side. Schöneberg 1938-1994 · Kreuzberg 1938-1994 · Reinickendorf 1974-1994 · Neukölln 1967-1992 (the whole holding) · Charlottenburg, Steglitz, Tempelhof, Tiergarten, Wedding, Wilmersdorf and Zehlendorf 1967-1994 — AND SPANDAU I 1983-1994 ONLY. She leaves the telephone book between the 1979-80 and 1980-81 volumes, so THE THREE UNREAD SPANDAU YEARS WERE HER THREE YEARS. Not a near miss: the shape of a sweep that counted boroughs and not years. ALL THREE ARE NOW READ — P Rep. 480 Nr. 1238 (1980), 1239 (1981), 1240 (1982), all LEGIBLE at 1.861, 1.725 and 1.691 control names a page — AND THERE IS NO KERGER IN ANY OF THEM, not once, and no Leibholz in 1980 or 1981. The forename is its own control: AGNES appears 90, 70 and 78 times across the three volumes, two hundred and thirty-eight other women, so the instrument plainly reads that word. SO: AGNES LEIBHOLZ NÉE KERGER DOES NOT APPEAR IN ANY WEST BERLIN DEATH REGISTER INDEX THAT EXISTS FOR HER WINDOW. Twelve boroughs, every year each one holds, 1967 or earlier to 1994. THREE READINGS REMAIN: she died outside Berlin (her husband's own case is the warning — Heinrich left the directory in 1942 and died in 1948 in a borough he did not live in); or after 1994, when she would be 105 and the thirty-year closure still hides it; or East Berlin, which is now bounded rather than shrugged at — LABSA holds eleven East offices, 605 volumes, 41.9 GB, and UNTIL NOVEMBER 1989 A WEST BERLIN PENSIONER COULD NOT RELOCATE EAST AT ALL, so it is live only for 1990-94 when she would be 101 to 105. In data/spandau-1980-1982-deaths.tsv. AND ONE PRECISION, BEFORE ANYBODY READS THE LIMIT THE WRONG WAY: the cross-indexing limit found in Spandau 1980 bites on a woman whose NAME OF RECORD AT DEATH is not the family's — Else Hedwig died a HERFERT. AGNES'S NAME OF RECORD IS LEIBHOLZ and the telephone book proves she used it, printed so at Mehringdamm 109 in every volume from 1972-73 to 1979-80. She would be entered “Leibholz geb. Kerger”, an L entry; and the twelve-borough sweeps read WHOLE VOLUMES by text layer anyway. THE SENTENCE ABOUT HER STANDS AT FULL STRENGTH. 11 OCTOBER, LATER — AND THE EAST BERLIN READING IS RETIRED AS FAR AS EVIDENCE CAN RETIRE IT. Scoping to the only years the hypothesis can live in — 1990, when the Wall had been open for weeks and she was 101, to 1994, when the index closes and she would be 105 — turned 605 volumes and 41.9 GB into 47 VOLUMES AND ONE GIGABYTE, all of which were downloaded and scored. THIRTY-THREE LEGIBLE AND READ, FOURTEEN BLIND AND NOT READ. No Leibholz, no Leipholz, no Leipsholz, no Luwinski and NO KERGER in any of the thirty-three — the first time the maiden name has been put to any East Berlin office. TWO OF THE ELEVEN BOROUGHS STOP AT 1990 REGARDLESS: Treptow and Hohenschönhausen have no 1991-94 index in LABSA at all, which is a holding limit and cannot be closed by reading. AND EVERY 1990 VOLUME IN EAST BERLIN IS BLIND, IN ELEVEN SEPARATE OFFICES — EVERY OFFICE IN EAST BERLIN — ALL AT 0.000 while their 1991 volumes score 1.4 to 2.5 — a scanning boundary at 1990/91, the THIRD this archive has measured after Kreuzberg's 1963 and Schöneberg's 1984. Prenzlauer Berg is blind for four years and only 1994 reads. THE FOURTEEN ARE NAMED IN THE FILE AND ARE NOT BEING READ, which is a judgement recorded as one: 1,759 pages, perhaps fourteen L-block reads, against a prior that requires a woman of 100 to cross a border shut her whole adult life and die in the east within five years — and an L-block reading would carry exactly the narrowness this archive spent the morning cautioning about. ONE KERGER CAME OUT, “Kerger, Karoline Henriette Lucie” at Köpenick 1992, AND SHE IS NOT AGNES. SO TWO READINGS REMAIN AND NEITHER IS REACHABLE TODAY: she died outside Berlin, or after 1994 behind the closure. In data/east-berlin-1990-1994-deaths.tsv.

Found on 21 September while sweeping Spandau for Erna, who is not there. "Leibholz verh. Schönfeld | [forename withheld] | 19. 5. | 1486 | mos. | verw." MOS. IS MOSAISCH. A JEWISH WOMAN BORN LEIBHOLZ, LIVING IN BERLIN-SPANDAU AND DYING THERE IN 1982. HER FORENAME IS WITHHELD AND THAT IS DAVID'S DECISION TO REVERSE, NOT THIS ARCHIVE'S. Set beside the surname she was born with, her forename is ALSO THE MAIDEN NAME OF A LIVING PERSON IN THIS FAMILY. The two are certainly different women -- one died in 1982 -- but a reader and a search engine cannot tell, and the rule here is that a living person's former name does not go on a public page whoever the page is about. The forename is in the private notes and the register number is published, so the certificate is orderable without it. THE FREE TEST BEFORE ANY PURCHASE: the Berlin marriage indexes for a LEIBHOLZ–SCHÖNFELD marriage, which names her father. TWO THINGS THIS ARCHIVE ALREADY HELD AND NEVER JOINED: Yad Vashem 5415605 gives SAMUEL LEIBHOLZ a wife with the same forename, recorded as new and never followed; and the Arolsen tracing files hold a Leibholz correspondence file INDEXED UNDER SCHÖNFELD -- the same married surname. THE TWO WOMEN ARE PROBABLY NOT ONE: Samuel's widow would be indexed under her own birth name as "Xxx verh. Leibholz", and this entry makes Leibholz the birth name. 23 SEPTEMBER — THE FREE TEST MAY NOT EXIST, AND THAT IS THE FINDING. This row calls the Berlin marriage indexes the free test before any purchase. THE MARRIAGE INDEXES STOP AT 1944, EVERYWHERE — 1,595 volumes and the latest year in any of them is 1944, the 80-year closure moving forward a year at a time. She was a Jewish woman in Berlin, widowed by 1982, who survived the war; A MARRIAGE IN THE LATE 1940s IS AT LEAST AS LIKELY AS ONE BEFORE 1944. So the free route settles it if the marriage is pre-1945 and CANNOT TOUCH IT otherwise. Worth knowing before 1,180 volumes are read to find nothing, and it changes when the purchase of Spandau 1982 Nr. 1486 stops being avoidable. THAT REMAINS DAVID'S DECISION AND NOT THIS ARCHIVE'S. 10 OCTOBER — THE “FREE MARRIAGE INDEX” TEST IN THIS ROW IS FREE BUT NOT SEARCHABLE. Four Berlin marriage volumes were sampled across two offices and three decades and ALL FOUR SCORE 0.0 CONTROL NAMES A PAGE — no extractable text whatever. The holding is 313 volumes and 28.3 GB to 1944. So the cost is not a download, it is an eye: unscoped, a surname is on the order of fifteen hundred image reads. THE INSTRUMENT IS NOT UNUSABLE, IT IS UNUSABLE UNSCOPED. In data/berlin-marriage-indexes-are-blind.tsv. This row had already found on 23 September that the indexes stop at 1944 and cannot touch a post-war marriage. That stands and is now worse: even inside the reachable years there is no office and no year to narrow by. 11 OCTOBER — RE-FOUND INDEPENDENTLY, AND SHE IS INDEXED TWICE. The Spandau 1982 volume was swept again from a different direction (the gap in row 253's coverage) and the entry came back: “Leibholz verh. Schönfeld · [forename withheld] · 19. 5. · 1486 · mos. · verw.” on PAGE 100 under L — AND “Schönfeld geb. Leibholz · [forename withheld] · 19. 5. · 1486 · mos. · verw.” on PAGE 168 under S. THE SAME REGISTER NUMBER FROM TWO UNRELATED POSITIONS, which is the pattern first seen with Gerda Bade geb. Leibholz at Wilmersdorf in 1993. THE FORENAME WAS READ AGAIN AND IS NOT WRITTEN DOWN AGAIN. It stays withheld for the reason this row gives, and that is David's decision to reverse. In data/spandau-1980-1982-deaths.tsv.

Commit 8c1d8af failed CI in 17 seconds and the site did not publish. THE GATE WAS RIGHT AND THE LOCAL RUN OF THE SAME GATE WAS WRONG. --harvest builds each guard from the live tree's first + last, which is a person's CURRENT name, and NEVER CHECKS A MARRIED WOMAN'S MAIDEN NAME. --hashes carries nine salted digests where harvest built seven guards, and the two extra are maiden-name forms. So a page naming a living woman by the surname she was born with passed locally and was stopped in CI. A GREEN LOCAL RUN HAD BEEN READ AS PERMISSION ALL MORNING AND IT WAS NOT ONE. FIXED: check:living now runs BOTH modes locally and the build fails if either does; the limit is written into the tool's own docstring beside the one already recorded for --hashes. PROVED: the hashes run locally now reproduces the CI failure name for name and file for file. A peer session found this by reading CI while this archive was reading its own green terminal.

DONE 21 SEPTEMBER, ALL OF IT. Twelve volumes, P Rep. 163 Nr. 584 to Nr. 595, read page by page at the image: 295 pages of L surnames, roughly 9,900 names. 1938 TO 1968 NOW RUNS WITHOUT A GAP, and with 1970-1994 already closed by text layer that is fifty-six years of Schöneberg death registration asked the question once and properly. THE RESULT IS ONE HIT IN THIRTY-ONE YEARS: Günther Leibholz, 1939 Nr. 780, which was found on the second day of the read and never repeated. Every year is bounded at both ends — a low opening register number and a closing entry in the last days of December followed by blank ruled lines — every page was read in two overlapping halves with the overlap confirmed by a shared row, and from 1954, when the register began grouping each letter-year by month, the headings were checked Januar to Dezember, all twelve, in order. READING ONLY THE L BLOCK IS ENOUGH, AND THAT IS PROVED RATHER THAN ASSUMED. A daughter born Leibholz who married would die under her husband's name, which would have put a hole in the negative exactly where this family's women are. The register files every married woman TWICE, under her husband's surname and under her maiden name, with the same register number — demonstrated with a named pair in EVERY ONE OF THE TWELVE VOLUMES, thirteen pairs in all. AND THE CLERKS COULD FORM THE LETTERS. The 1967 L block contains LINDEHOLZ (Nr. 268), which is how these hands write -holz, and LEIPOLD (Nr. 1290), which is how they write Leip-. Both halves of the name occur in Schöneberg. Only the combination does not. Fourteen names had to be settled at scale 6, the two closest being LIEBHOLD (1958 Nr. 2136) and LEITHOLD (1963 Nr. 734, again 1968 Nr. 2366) — each one letter from Leibholz in the middle and one at the end, and each indistinguishable from it at page scale. The full method and every page range is in data/schoeneberg-progress.tsv, which is the control file any future session reads first.

The first and only Leibholz in eleven years of reading Schöneberg's death indexes at the image. THE ENTRY SAYS "ev." AND CARRIES NO "ISRAEL", AND IN 1939 THAT IS EVIDENCE RATHER THAN SILENCE. From 1 January 1939 Jewish men were compelled to add ISRAEL, and this register applies the rule relentlessly — the same nine pages carry Lehmann Bernhard Israel, Löffler Gustav Israel, Litten Minna Sara, Lichtheim Berta Anna Sara and dozens more, all marked mos. Günther carries neither. THE STATE DID NOT CLASSIFY HIM AS JEWISH. That damages the obvious hypothesis: Jacob Leibholz was mos. and his wife Auguste was kath., so a child of theirs would be Catholic or Jewish, NOT PROTESTANT. NOT KIN ON ANY EVIDENCE HELD. TESTS IN ORDER OF COST, AND THE FIRST TWO ARE FREE: (1) the 1939 and 1940 Berliner Adressbuch for a Leibholz household in Schöneberg — this archive's 1939 list has Ella, Fritz and Jaques there and no Günther; (2) the Berlin MARRIAGE indexes, which are open to 1938, for a Leibholz marrying a Protestant wife, since births are closed 110 years and a Günther of the 1920s or 1930s is unreachable; (3) the certificate itself, Schöneberg 1939 Nr. 780, which carries the age, the address, the parents if he was a minor, and the informant. THE CERTIFICATE IS A PURCHASE AND IS NOT MINE, and it is the cheapest document in the whole backlog that could turn a name into a person. 22 SEPTEMBER — TEST (2) COSTED AT THE IMAGE, AND THE WORD “FREE” IN THIS ROW IS DOING TOO MUCH WORK. The full inventory now exists for the first time: TWENTY-ONE Schöneberg marriage name-indexes across three offices, covering 1874 to 1943 without a gap — P Rep. 160 (Schöneberg I, ten volumes), P Rep. 161 (Schöneberg II, seven), P Rep. 163 (the merged office from 1938, four, of which 1938-39 IS SPLIT INTO SEPARATE MEN'S AND WOMEN'S BOOKS, the only place in the run where that happens and a trap for anyone who reads one and thinks they have read the year). The archive had previously named two of the twenty-one. THEN THE STRUCTURE WAS MEASURED ON P REP. 161 NR. 270 AND IT KILLS THE CHEAP VERSION OF THIS TEST. It is ONE ALPHABET with the years NESTED INSIDE EACH LETTER — the same shape the 1949-51 death indexes turned out to have — so you cannot open a volume at the year you want. AND WITHIN A LETTER THE ENTRIES ARE IN REGISTRATION ORDER, NOT ALPHABETICAL: page 256 of the L block reads Lintow, Lungwitz, Liebermann, Landois, Loewenthal, Lehmann. The first letter is sorted and nothing else is, SO THERE IS NO SKIPPING TO “LEI” — every page of every L block must be read. L occupies about forty-two pages of that four-year volume, the Kurrent hand needs scale 2.6 to be legible, and at that scale a page is THREE READS. About 125 reads a volume; sixteen volumes cover 1899-1938; ROUGHLY TWO THOUSAND READS, nearly ten times the Kreuzberg backlog in row 253. SO THE ORDER OF THE THREE TESTS IN THIS ROW IS WRONG AND IS HEREBY REVERSED. The certificate — Schöneberg 1939 Nr. 780 — is ONE DOCUMENT carrying his age, his address, his parents if he was a minor, and the informant, and it would settle in a page what two thousand reads might not settle at all, because a marriage index yields a surname and a register number and says nothing about a son. AND THE WHOLE ROUTE RESTS ON AN UNASKED ASSUMPTION: that his parents married in Schöneberg. He DIED there. Berlin had well over a hundred registry offices, and an empty Schöneberg sweep would prove almost nothing. THE INVENTORY AND THE ARITHMETIC ARE IN schoeneberg-marriage-indexes.tsv. THE ROW STAYS OPEN and now recommends the purchase over the sweep. TEST (1) COMPLETED FOR 1940, 22 SEPTEMBER. The row had the 1939 list already (Ella, Fritz and Jaques in Schöneberg, no Günther); 1940 was outstanding. The 1940 Berliner Adressbuch is in the ZLB full-text index — controlled by “Schmidt” returning 3,106 in that volume — and LEIBHOLZ RETURNS TWELVE, LEIPHOLZ SIX. The twelve, as the street sections print them (initials only, which is how the Straßenteil works): S. Israel · A. Israel · W. Israel · F., Pensionär · V., Witwe · one read as Bankbeamter · SL., Witwe · F., Asm. · M., Witwe · H., Verwalt. — plus the two cross-reference lines “Leibholz s. a. Leipholz”. THREE OF THEM CARRY “ISRAEL”, the forename compulsory for Jewish men from January 1939 — the same rule whose ABSENCE from Günther's death entry is this row's whole argument. AND THERE IS NO G. AMONG THE TWELVE. That is consistent with a man who died in Schöneberg in 1939 and would not be in the 1940 book — it is what the row expects rather than a discovery, and it is recorded because the row asked for it. TWO LIMITS: the OCR of this volume is poor, rendering initials as “tz.” and “SL”, so a G could be mangled; and the street sections list HOUSEHOLDERS, so a son living at home would not appear under his own name whether he were alive or dead. 10 OCTOBER — THE “FREE MARRIAGE INDEX” TEST IN THIS ROW IS FREE BUT NOT SEARCHABLE. Four Berlin marriage volumes were sampled across two offices and three decades and ALL FOUR SCORE 0.0 CONTROL NAMES A PAGE — no extractable text whatever. The holding is 313 volumes and 28.3 GB to 1944. So the cost is not a download, it is an eye: unscoped, a surname is on the order of fifteen hundred image reads. THE INSTRUMENT IS NOT UNUSABLE, IT IS UNUSABLE UNSCOPED. In data/berlin-marriage-indexes-are-blind.tsv. This row's own last line already said the word “free” was doing too much work. IT NOW HAS A NUMBER: zero extractable text, 313 volumes, and no narrowing available for a man whose age, address and parents are all unknown.

OPENED AND CLOSED 21 SEPTEMBER. Dumping every 'person' the evidence gate knew about showed that about a fifth were not people but ROW LABELS out of the narrative TSVs — ' AND A CHILD THIS ARCHIVE DID NOT HAVE ', 'AFRIKANISCHE STRASSE 37 — and a namesake to keep apart', 'who she was looking for (1)'. They arrive through the manifest's ["who"] key, because in a narrative file that column carries section headings as well as names. TWO COSTS, AND THE SECOND IS THE REAL ONE. The published figure overstated: /register/ and the evidence panels said 187 people and 252 records. And for those rows check-evidence's promise quietly degraded from 'this PERSON reaches a reader' to 'this STRING is somewhere on the site' — which it trivially is, because the sentence is printed on the page. THOSE ROWS WERE PASSING THE GATE FOR THE WRONG REASON. THE FIX IS RECOVERY, NOT REJECTION. The rows are written 'NAME — what the row is about', so build-evidence now cuts at the em dash, at ' (', and at a comma WHOSE TAIL IS PROSE — the last so that 'ARTHUR, THE SANITÄTSRAT' yields ARTHUR while 'Leibholz, Samuel' is left alone. What survives must then still look like a name: no digits, no leading asterisk or bracket, no opening lower case, no more than six words, no sentence words. THAT RECOVERED TWELVE PEOPLE WHO HAD BEEN STORED UNDER A WHOLE SENTENCE — HANS LEIBHOLZ, HEINRICH LEIBHOLZ, ANTON LUWINSKI, SOPHIE LEIBHOLZ, AUGUSTE LEIBHOLZ née LUWINSKI, IVONE LUWENSKI, PAULINE BERMAN, PAULINE MILLER and others — whose names were previously 'HANS LEIBHOLZ — read in full' and 'PAULINE BERMAN (formerly MILLER, born RAPPOPORT)'. NOTHING IS DROPPED SILENTLY. build-evidence now PRINTS every cell it refused, 52 of them, with its file, under the line 'if a real name is in this list, the data or the rule is wrong'. A filter you cannot see is the fault this archive keeps having, and the first draft of this rule wrongly rejected IVONE LUWENSKI, ARTHUR and JACOB — the printed list is how that was caught and corrected in one pass. VERIFIED BEFORE AND AFTER. The old and new people lists were diffed name by name and every real person was confirmed still present: George William Castle, Heinrich Leibholz, Hans Leibholz, Jacob Leibholz, Samuel Israel Leibholz, Oskar Leibholz, Auguste, Kurt, Sophie, Anton. THE PUBLISHED COUNT IS NOW 204 RECORDS ABOUT 136 PEOPLE, and it is the first time that number has meant what it says.

Row 248 asked for three things — every atlas pin, every stated distance, every date converted between a register and a narrative. THE PINS AND THE DISTANCES ARE DONE AND FOUND TWO WRONG PINS; THE DATES ARE NOT STARTED, and saying so in its own row is better than letting a third of a task sleep inside a green one. WHAT IT MEANS IN PRACTICE: every date this archive prints in prose should be reconcilable to the register entry or document it came from, and nobody has ever walked that list. The known shape of the fault is the Sauerbaum pin and the 'about fifteen kilometres' — a figure that was close enough to survive a published page, a build gate and several re-readings, and was caught only because something unrelated touched it. A date that is a year out will behave exactly the same way. START WITH DATES THAT APPEAR IN BOTH A data/ FILE AND AN .astro PAGE, because those can be compared mechanically, and only then read the ones that exist in prose alone. 22 SEPTEMBER — FIRST PASS RUN, AND IT FOUND ONE. A mechanical audit of all 149 register people (death before birth, born/died disagreeing with the derived by/dy, facts dated outside a life, impossible calendar dates, a child born before its parent was thirteen) returned THREE hits, of which one is real and published. KURT ISRAEL LUWINSKI READ 'died 15 December 1968' AND 'dy 1969' IN THE SAME RECORD. The archive corrected his death from Dec 15 1969 to 15 December 1968 on good evidence — the Johannesburg Chevrah Kadisha burial register, died 15.12.1968, buried 16.12.1968, Westpark block L grave 211, Yahrzeit 24 Kislev 5729, which is itself 15 December 1968, and burial the next day is what you would expect. THE CORRECTION WAS RIGHT AND IT DID NOT PROPAGATE. build-register.py applied it to died and never recomputed dy, and dy is what /bloodline/ prints and what /timeline/ PLOTS HIM BY. So the register page said 1968, the bloodline said 1912-1969, and the timeline put him in the wrong year. Two prose lines had gone stale with it — data/berlin-postwar-luwinski.tsv and a research-log entry both still read 'died in Johannesburg in 1969'. FIXED: the builder now recomputes dy from a died correction, and does the same for born/by, which was not triggered by the data either — and neither was the died case when it was written, and it still shipped a wrong year onto two pages. The two prose lines are corrected. The site now carries one death year for Kurt in every place, and the only 1969s left are Henry Robert Wear, who did die in Johannesburg in 1969, and the correction's own record of what it superseded. THE OTHER TWO HITS ARE BENIGN — a Residence dated to the year of birth, and a stray fact on the same record. STILL OWED: the 52 date strings the parser cannot read ('Circa 1912', 'Between Aug 1966 and July 1970', '25. o. 26 Mrz. 1871') are unaudited, and no date in prose has been reconciled to its source document. THIS PASS CHECKED THE REGISTER AGAINST ITSELF, WHICH IS THE CHEAPEST THIRD OF THE JOB. 22 SEPTEMBER, SECOND LEG DONE — AND THE ROW'S OWN NUMBER WAS WRONG AGAIN. This row said 52 date strings the parser cannot read. EVERY DATE STRING IN THE REGISTER PARSES; not one fails. The 52 counted strings that are not in the canonical shape, and most of those are just “Aug 8 1884”. THE REAL POPULATION IS THIRTY-ONE STRINGS CARRYING A QUALIFIER, and the danger was never that they parse badly — IT IS THAT THEY PARSE TOO WELL. year() takes the first four digits and throws the qualifying word away, so “Before 1955” reached the bloodline as 1955 and the timeline PLOTTED A POINT AT IT. A BOUND WAS BEING PUBLISHED AS A DATE, IN BOTH DIRECTIONS. Twenty-seven of the thirty-one are on facts, which print as the source wrote them and are safe. FOUR REACHED A READER AS BARE YEARS: Joachim Luwinski “Before 1955”, Elli Leibholz (Struzik) “After May 13 1942”, Johannes Antonie Oelofse De Witt “Circa 1923” and Maria Elizabeth van Straaten “Circa Jan 1876”. FIXED IN THE BUILDER, WHICH IS THE LESSON FROM THE KURT CASE FOUR HOURS EARLIER: build-register.py now has qualifier() and every record carries byq/dyq beside by/dy, the correction loop sets them too so a correction cannot re-introduce the fault, and both pages print the word — the bloodline reads “1865 – before 1955” and the timeline says in its date column what kind of point it is. AND ONE OF THE FOUR WAS REPLACED OUTRIGHT BY THE SAME EVENING'S OTHER WORK: Elli's “After May 13 1942” is the DEPORTATION DATE, and Gedenkbuch entry en1169257 — loaded into this archive hours before — gives her as dying at KULMHOF IN MAY 1942, birth date matching to the day. Correction 9500147. THE YEAR ON THE PAGE DOES NOT CHANGE, and that is worth saying plainly: the archive was right by accident, because a floor and the truth fell in the same year. Joachim's bound has no such luck behind it. The whole audit is in register-date-audit.tsv. THE ROW STAYS OPEN ON THE THIRD LEG, which is the expensive one: not one date printed in prose has been reconciled to the document it came from. Both passes so far have checked the register against itself. THIRD LEG, MECHANICAL PART DONE 22 SEPTEMBER, AND IT COMES BACK CLEAN. *** 156 distinct full prose dates across the pages; each taken with 160 characters of context; every published person with a known year looked for in that context; any date within two years of their register date but not equal to it flagged. ELEVEN FLAGS, ALL ELEVEN FALSE POSITIVES, each opened and read — every one is a date about one person sitting near another person's name. NOT ONE PROSE DATE CONTRADICTS THE REGISTER. AND THE CHECK PROVES ITS OWN LIMIT: proximity is not aboutness, so it can show that no prose date disagrees with a named person's register entry, but it CANNOT reconcile a prose date to the DOCUMENT it came from, which is what this row asks. That part needs a reader with the source in front of them, one date at a time. THE ROW STAYS OPEN ON THAT, and is now closed on everything a script can do.

Row 246 expected the Gedenkbuch to hand over Berlin ADDRESSES and it hands over BOROUGHS. Forty-four of the fifty-two are Berlin — Schöneberg 6, Wilmersdorf 6, Charlottenburg 5, Friedrichshain 5, Tiergarten 4, Neukölln 4, Mitte 4, Prenzlauer Berg 3, Kreuzberg 1, three plain Berlin, two with two boroughs each. THE WORK IS THE REVERSE OF WHAT ROW 246 IMAGINED: not look up an address, but read the Adressbuch's Leibholz entries for 1933 to 1943 and match each one to a person by borough, forename and household. The Adressbuch prints street and borough, so a borough plus a forename is usually enough to decide. WHAT IT WOULD BUY: a trade and a street for each of forty-four people, and — for the fourteen who appear only under a married surname — the chance to find the household they left. THE TRAP TO AVOID IS THE ONE ROW 246 FELL INTO: a borough is a filter, not an identifier, and two Leibholz in Charlottenburg are two Leibholz in Charlottenburg until something else separates them. 22 SEPTEMBER — THE ADRESSBUCH HALF MEASURED, AND FOUR OF THE ELEVEN YEARS CANNOT BE DONE THIS WAY. The householder runs were read AS RUNS rather than by name search, which is the instrument row 251 arrived at the same morning: find the headword, take every dashed line under it, stop at the next surname. IT REPRODUCES THE IMAGE READING WHERE THE DASHES SURVIVE — 1938 returns exactly seventeen and 1940 exactly ten, matching what was read off the page images on 20 September. IT FAILS WHERE THEY DO NOT: 1935, 1937, 1939 and 1941 return no run at all because the headword is mangled past matching, and 1942 and 1943 return one and two entries against true counts of six and four, because the lost em-dashes swallow the next surname as a continuation. SO THE BOROUGH MATCHING NEEDS THE PAGE IMAGES FOR 1939, 1941, 1942 AND 1943 — which this row always implied and is now measured. The pass paid for itself twice on the way: THE 1940 COUNT WAS ELEVEN IN THREE FILES AND IS TEN (the eleven belongs to 1939), and JAQUES LEIBHOLZ & CO. WAS AT KRAUSENSTRASSE 59-60, NOT KRONENSTRASSE — four readings in three volumes, and the long-s artefact in "Kraufen" and "Kraujeu" is a letter Kronenstrasse does not have. Corrections 9500911 and 9500912. Working in data/adressbuch-namenteil-runs.tsv. 11 OCTOBER — THE ROW WAS BLOCKED ON WORK THAT WAS ALREADY FINISHED, AND IT IS NOW DONE. The 22 September conclusion above — “the borough matching needs the page images for 1939, 1941, 1942 and 1943” — WAS WRONG. Those four volumes had been read at the image on 20 SEPTEMBER, two days earlier, and the per-year lists were already in data/berlin-adressbuch-names.tsv. The row sat still for nineteen days waiting for work it already had. All four were re-read at the image anyway, independently, which is how that was established: ELEVEN, NINE, SIX AND FOUR, every forename agreeing with the 20 September reading and none in one list and not the other. WHAT THE RE-READING ADDED, BEYOND THE REPLICATION: the 1939 trades and street numbers, which the old list did not carry, and with them THREE CHANGES BETWEEN 1938 AND 1939 — Albert from Antiquitäten to Möbel at the same house; PETER from Fabrikdirektor at Drakestraße 1 to FABRIKBESITZER at Grunewald, a move and a promotion in 1939, which is the wrong direction for a Jewish businessman that year; and WILHELM OF WEINSTRASSE 6 from Kaufmann to VERTRETER, which moves the start of his demotion a year earlier than this archive said and makes its first step an unmarked one (correction 9500918). AND THE MATCHING ITSELF, WITH ITS YIELD COUNTED HONESTLY. FORTY-SEVEN Berlin Gedenkbuch people, not forty-four — the row's own figures never added up to its own total (correction 9500917). Of the forty-seven the method ATTACHES FOUR to one printed household (Else née Seelig, Kurt, Lieselotte and Walter to “A Israel, Fleischermeister, NW87 LEVETZOWSTRASSE 13”, who becomes “A Israel, Arbeiter” at the same address in 1941), CORROBORATES THREE already attached (Georg, Lydia and Elli of Michaelkirchstraße 26 — one borough, one transport, one death date), CONFIRMS ONE already known (Samuel at Kommandantenstraße 29, the only borough where the arithmetic forces it: one Gedenkbuch person in Kreuzberg and one household), and PUTS A DIRECTION ON ONE unordered pair of boroughs (Hans: Tiergarten 1938, then Charlottenburg 1939). IT DOES NOT TOUCH THE OTHER THIRTY-EIGHT, and thirty-eight is the number that matters. THE TRAP THE ROW WARNED ABOUT WAS SPRUNG, IN NEUKOLLN, AND WAS NOT CAUGHT BY THE BOROUGH. Two Leibholz households there and only one in the Gedenkbuch: the butcher's, and Heinrich of Leinestraße 32, who carries no “Israel” in any volume and is still printed in 1942. The trade and the marking separated them; the borough could not. AND CHARLOTTENBURG HOLDS FIVE OF THE DEAD AND NOT ONE PRINTED HOUSEHOLD, Prenzlauer Berg three and none — people who lived in somebody else's flat are not householders and are not in the Namenteil, which is the shape of the source and not a failure of the reading. In data/adressbuch-boroughs-1938-1943.tsv.

DONE 22 September 2026, and it never needed a PDF viewer. The chunk's OCR text is served as a property, textract:docContent, so the page is never rendered at all — the CSP and the hidden-pane problem that blocked this an hour ago were both beside the point. FOUR CHUNKS READ IN FULL: 009223_weake_1973.pdf, which runs Weake to Weare and therefore holds the whole of WEAR; 009474_wear_1974.pdf; 008820_weale_1979.pdf; 009151_wear_1980.pdf. THE TWO FOLLOWING YEARS WERE READ BECAUSE THE CALENDAR RUNS BY YEAR OF GRANT, NOT YEAR OF DEATH — the 1973 volume contains people who died in December 1972, which is the trap this row was opened warning about. SIXTEEN WEAR GRANTS, COUNTED OFF THE PAGE: 1973 Alfred, Dorothy Winifred, Ellen, Elsie, Ernest Leslie, James Watson, Mary Rosine, William; 1974 Walter; 1979 James William, Arthur, Charles, Edith Maud; 1980 Annie Ellen, Norman, Rosetta Emma. NONE IS HIM. AND THE WIDER TEST IS WORTH MORE THAN THE NARROW ONE: the forename pair “Leonard Aubrey” put to every year from 1970 to 1981 returns TWENTY-SIX MEN and not one is a Wear — Bricknell, Dinsmore, Williams, Day, Hill, Beniston, Moore, Barton, Peters, Waghorne, Evans, Hogben, Pridole, Webberley, Howard, Briers four times, Cooper, Hillier, Palmer, Hawkins, Wines, Bright, Stuckey. CONTROLS WERE RUN BEFORE ANYTHING WAS BELIEVED, which is the whole difference from the three failed attempts: “Ernest Leslie” 1973 returns 18, the nonsense string returns 0, and WEAR/SMITH/LUWINSKI return 41/469/0. WHAT IT DOES NOT SETTLE: an English grant needs English assets, so a man who died in South Africa need never appear here. This closes a route, not a life. Written up in probate-calendar-wear.tsv. Nothing downloaded, nothing ordered.

DONE 22 September 2026, within the hour of the Bydgoszcz letter arriving. STANDESAMT SCHUBIN, GEBURTSREGISTER 1884, AKT NR. 63. Registered at Schubin on 11 AUGUST 1884: “der Kaufmann SALLY LEIBHOLZ”, resident at Schubin, mosaischer Religion, reported that of “FRIEDERIKE LEIBHOLZ GEBORENE SCHMUL”, his wife, mosaischer Religion, living with him at Schubin, on the EIGHTH OF AUGUST 1884 a child of male sex was born, given the forename JACOB. SALLY SIGNED THE ENTRY HIMSELF. USC Szubin Miasto, sygn. 28, scan 64, modal id 295772956. WHAT IS NEW IS NOT THE PARENTS. The archive already held all three people and the link between them from the family record. WHAT CHANGES IS THE STANDING: Jacob's parentage rested on a tree this morning and rests on the civil register of his birthplace tonight, and the two agree to the day. Plus five things the tree did not have — Sally's trade (Kaufmann) dated to 1884, both parents' religion as the STATE recorded it, the registration date, the shared residence, and the first specimen of Sally's handwriting this archive has. AND IT DOES NOT ANSWER THE QUESTION THE LETTER WAS SENT TO ASK. A German birth entry names the parents and NOT THEIR ORIGINS — no birthplace, no age, no former residence for either of them. So whether the Schubin Leibholz are a Hammerstein family that moved east is exactly where it was: on the Gedenkbuch, and on a Hammerstein birth in 1847 that Koszalin has confirmed no surviving register can now verify. THE NEXT DOCUMENT IS THE ONE THAT WOULD SETTLE IT — see row 263.

DONE 22 September 2026, the same evening it was opened. STANDESAMT SCHUBIN, AKT NR. 9 OF 24 NOVEMBER 1874 — the ninth and last marriage of the year. DER KAUFMANN SALLY LEIBHOLZ, mosaischer Religion, SIEBEN UND ZWANZIG JAHRE ALT, GEBOREN ZU HAMMERSTEIN, WOHNHAFT ZU HAMMERSTEIN, son of the Kaufmann JACOB LEIBHOLZ and his wife WILHELMINE GEB. POLLNOW; married FRIEDERICA SCHMUL, mosaischer Religion, geboren zu Schubin, wohnhaft zu Schubin. THIS ANSWERS THE STANDING QUESTION FROM A DOCUMENT. The Schubin branch is a Hammerstein family that moved east, and it is no longer an inference from the Gedenkbuch — a registrar wrote it down in 1874, and twenty-seven years old in November 1874 puts his birth in 1847, the year this archive has always carried and which Koszalin confirmed no surviving register can verify. THE MARRIAGE ENTRY DID WHAT THE LOST BIRTH REGISTER WOULD HAVE DONE. AND IT ADDS A GENERATION: JACOB LEIBHOLZ and WILHELMINE NÉE POLLNOW, Sally's parents and Kurt's great-grandparents, neither of whom existed in this archive in any form before tonight. It also explains a name — the Jacob born at Schubin on 8 August 1884 was named for his grandfather. THE MAIDEN NAME IS THE LEAST CERTAIN THING ON THE PAGE and is flagged as such: Kurrent P/T/Z confuse, and it is read as POLLNOW. IT DOES NOT MAKE GEORG A BROTHER — two births two months apart in different provinces still cannot be brothers — but the reason for expecting the Hammerstein Leibholz to be strangers has gone. AND ONE MISTAKE ON THE WAY, RECORDED: “page 10” is SCAN 10, not scan index 9. The first attempt landed on akt Nr. 8, then stepped back to Nr. 7 — 24 November 1874, THE RIGHT DATE AND THE WRONG PEOPLE, which is precisely the near-miss that gets published as a find. The unit's own “Indexes in unit” tab settled it in plain text. THE RULE: when a viewer offers an index, read the index before paging by arithmetic.

He has no death date and no death place. Alive at Schubin in August 1884, born 1847, and that is the whole of it. THE SZUBIN ROUTE HAS BEEN COSTED AND IT IS BAD VALUE FIRST: 190 death units across the town's THREE registry offices, NO alphabetical death index (birth and marriage skorowidze exist; death ones return zero against a working control), and the portal's person index holds ONE Leibholz in 228,278 people. No index, no year, 190 units. TRY BERLIN FIRST. Kurt was born there in 1912, the household is at Chodowieckistraße in 1913, Jacob married there in 1924 — if Sally followed his sons he died in Berlin, and the Landesarchiv Namensverzeichnisse are an instrument this archive already reads fluently and has swept 142 offices with. A man born in 1847 dying in Berlin lands in a range the death indexes cover. IF BERLIN COMES BACK EMPTY, the Szubin sweep is still there and should start with USC SZUBIN MIASTO (6/1792), which is this family's office — not the two Wieś offices. PROBED 22 SEPTEMBER, AND THE CHEAP TEST CAME FIRST. Before sweeping any death index, the question is whether Sally was ever in Berlin at all — and the digitised Berlin address books are full-text searchable. The books print surname first, so the phrase had to be built that way. RESULTS: “Leibholz Georg”, the control, returns 4. “Leibholz Sally” returns 0. “Leibholz Salomon” returns 0. The null string “Leibholz Zqxvwyrk” returns 0. AND THE CONTROL IS WHAT MAKES THIS WEAK RATHER THAN STRONG, WHICH IS WHY IT IS WORTH STATING. FOUR HITS FOR GEORG IS FAR TOO FEW. He is in the Berlin directories across years — this archive has him at Chodowieckistraße 31 in 1913 and 1914 and at a Wäscherei in Michaelkirchstraße later. A phrase index that returns four for a man who should return dozens is SPARSE, so a zero from it is a weak negative and nothing more. It is recorded because it lowers the prior on Berlin slightly and cost three queries, NOT because it closes anything. Fraktur OCR mangles forenames, the digitisation does not cover every volume, and he may simply have died at Schubin before the family ever moved. 27 SEPTEMBER — ONE WAY OF MAKING THIS CHEAP WAS TESTED AND DID NOT CONFIRM. This row costs the Szubin route at 190 units BECAUSE THERE IS NO ALPHABETICAL DEATH INDEX, and that negative is about SEPARATE skorowidz units, which is sound. BUT PRUSSIAN ANNUAL VOLUMES ORDINARILY BIND A NAMENSVERZEICHNIS AT THE BACK OF THE BOOK ITSELF, uncatalogued — which is exactly how this archive reads Berlin. If Szubin does that, the sweep is not 190 units of paging but 190 LAST PAGES and the L slot on each, which is an afternoon rather than a project. TESTED ON THE ONE VOLUME THIS ARCHIVE CAN ALREADY OPEN: unit 6/1792/0/2.2/141, the 1874 marriage book, has twelve scans and SCAN 11 IS A BLANK UNUSED PRINTED FORM. No index. AND THE TEST IS WEAK AND IS RECORDED AS WEAK: that is the worst possible volume to try it on — registration began 1 October 1874, so the book is a THREE-MONTH STUB OF NINE ENTRIES, the volume least likely in the series to carry an index. NOT SUPPORTED, NOT REFUTED. The next test is the last scans of one FULL ANNUAL DEATH VOLUME of USC Szubin Miasto. WHAT BLOCKED THAT, so nobody repeats the hour: THE PORTAL WILL NOT OPEN A UNIT BY ITS SIGNATURE — query.type=UNITS is silently ignored, a scan's modal id resolves ONLY inside the current search result set so the neighbouring unit cannot be guessed, and the Vaadin search box could not be re-driven with a modal open. The one route that works reaches any scan of a unit you are already inside and no further: open a scan's modal, click VIEW THE DOCUMENT for the unit's scan list, click Show, and harvest the new Loris id from the network log. See data/szubin-death-registers-searched.tsv. 27 SEPTEMBER 2026 — THE PREMISE OF THIS ROW'S COSTING WAS WRONG, AND THE ARCHIVE ITSELF CORRECTED IT. This row costed Szubin at "190 death units, NO alphabetical death index, no year" and sent the reader to Berlin first on that basis. THE INDEX EXISTS. Archiwum Państwowe w Bydgoszczy sent USC SZUBIN MIASTO SYGN. 246 on 25 September under case IU.631.434.2026 — "Alphabetisches Verzeichnis zu dem Sterbe-Register über Standes-Amt Schubin von 1894 bis 1908", 103 scans — IN ANSWER TO A LETTER THAT ASKED ABOUT SOMETHING ELSE. The zero this row relied on was a fact about the PORTAL and was turned into a fact about the HOLDINGS. A catalogue is not an archive. THE L SLOT IS READ OUT AND HE IS NOT IN IT. K ends scan 44, L runs scans 45-48, M begins scan 49, every year 1894-1911 present. Lehmann, Lis, Luckstaedt, Lewandowska, Liesske, Lurkiewicz, Luckowski, Lesner, Luczny, Leonski, Lawnicki, Lauge, Lawrenz, Lesniewski, Lukowska, Lüttke, Luchowska, Lichy, Ludes, Lukomski, Loewe, Langer, Loeffler, Lölke, Leu, Lahren, Leiser, Lorusch, Łowiński, Loferski, Lange — NO LEIBHOLZ. AND THE CONTROL IS IN THE SLOT: the first entry is LEHMANN MOSES, Schneider, and 1909 opens with LEISER ADOLF; Samuel, Silberberg, Salzbrunn and Schobelt are elsewhere in the volume. THE BOOK LISTS JEWS, so the absence is real. SCHMUL WAS READ IN THE SAME SITTING — S and Sch, scans 80-88, no Schmul — so Juda Schmul, Jette Machot and Friederica did not die at Szubin town in the window either. WHAT IS STILL OPEN, AND IT IS THE LIKELIER WINDOW: THE VOLUME BEGINS IN 1894 and Sally was alive in August 1884 and born 1847, so 1884-1893 is entirely uncovered and he would have been only 46 in 1894. It is also ONE OFFICE OF THREE — Szubin Wieś I and Wieś II are untouched. NEXT: ask Bydgoszcz whether an equivalent Verzeichnis survives for 1874-1893. See data/szubin-sterbe-index-1894-1908.tsv. 11 OCTOBER — BERLIN WAS TRIED FIRST, AS THIS ROW ASKED, AND IT IS EMPTY. THERE IS NO SALLY LEIBHOLZ IN ANY BERLIN DIRECTORY FROM 1903 TO 1920. The Leibholz Namenteil run was pulled for 1903, 1905, 1908, 1910, 1913, 1916 and 1920 — volumes that print FULL FORENAMES, not initials — and the eighteen years yield Adolf, Albert, Arthur, Cäsar, Curt, Else, Felix, Franz, Georg, Ida, Julie, Julius, Leo, Louis, Max, Nathan, Oscar, Paul, Siegfried, Th. and William. NOT ONE SALLY AND NOT ONE SALOMON. The 1916 run was READ AT THE PAGE IMAGE and agrees with its text layer entry for entry, which is what licenses the other six years. STATED NARROWLY: a man who is not a householder is not in the directory, so this is not proof he never came to Berlin — he was 56 in 1903 and 73 in 1920 and could have lived with a son. What it is: the only cheap instrument for placing him in Berlin has been run across eighteen years and returns nothing, and THIS ROW'S “TRY BERLIN FIRST” NO LONGER HAS ANYTHING BEHIND IT. AND THE INSTRUMENT HAD A HOLE NOBODY HAD NOTICED: the ZLB anchor this archive has used for three weeks, 34115495, IS 1903 TO 1943 AND NOT 1799 TO 1943 — 1902 and earlier return zero pages and zero Schmidt, so it is the anchor and not the surname. The earlier volumes are under PI 34115512, 1873 TO 1896, now tested year by year with controls; AND 1897 TO 1902 ARE NOT DIGITISED UNDER EITHER, a six-year hole. The pre-1900 volumes give FORENAME INITIALS ONLY and cannot be asked a forename question at all. SZUBIN IS UNCHANGED AND STILL BAD VALUE; what has changed is that the alternative is tested and empty, and A DEATH AT SCHUBIN BEFORE HIS SONS LEFT remains the simplest explanation of every negative held. In data/berlin-directories-before-1903.tsv.

THIS ROW WAS WRONG WHEN IT WAS WRITTEN AN HOUR AGO AND IS CORRECTED HERE RATHER THAN QUIETLY REDONE. It said 119 indexed Schmul scans were waiting to be opened, implying a rich Schubin Schmul archive. THE 119 IS REAL BUT IT IS REGION-WIDE. FILTERED TO SZUBIN IT IS TWO, and one of the two is the 1874 marriage itself. The other 117 are Schmuls elsewhere in Kuyavia-Pomerania with nothing yet joining them to this family. The contrast that IS real and does stand: 119 Schmul against ONE Leibholz in a 228,278-person index. THE BRIDE'S HALF OF THE MARRIAGE HAS NOW BEEN READ. FRIEDERICA SCHMUL, jüdischer Religion, twenty-something, GEBOREN ZU SCHUBIN, WOHNHAFT ZU SCHUBIN. AND THE PRINTED FORM SAYS “TOCHTER DER” WHERE THE GROOM'S SAYS “SOHN DES” — same page, same hand. That is either “Tochter der unverehelichten [name]”, which would make Schmul her MOTHER'S name with no father recorded, or “Tochter der Eheleute X und Y”, which is equally standard. THE WORD AFTER “der” WAS NOT LEGIBLE AND IS NOT BEING GUESSED. No parent of Friederica has been entered into this archive. THE ONE REAL LEAD: ABRAHAM SCHMUL, born Szubin 1846, died Koronowo 1904, son of JULIUSZ SCHMUL and JETTE NÉE MUSCHILL. A Jewish Schmul born in the same small town six to eight years before her. NOT CLAIMED AS HER BROTHER — no document joins them and the card is a volunteer transcription, not a register. AND THE TRAP THAT WAS AVOIDED IS WORTH THE ROW ON ITS OWN: Abraham's card names a MARRIED couple, which would make “Tochter der Eheleute” suddenly readable. A hypothesis that makes a hard word legible is the oldest way to misread a document, and this archive did exactly that twice today. AND THE NEXT STEP THIS ROW ORIGINALLY PROPOSED WAS IMPOSSIBLE, WHICH WAS CHECKED RATHER THAN ASSUMED. It said to read Friederica's own birth entry in the Szubin birth books “which are online and cover the 1850s”. THEY DO NOT AND CANNOT. PRUSSIAN CIVIL REGISTRATION BEGAN ON 1 OCTOBER 1874 — which is why the whole 1874 marriage unit holds NINE entries — so a birth around 1852 has no Standesamt record anywhere in Prussia. A Jewish birth of that decade could only be in the records of the Jewish religious community, and the portal's team facet, probed with controls, has none: “Miasto” returns six registry offices including Szubin Miasto, “Gmina” returns exactly one religious fonds — the EVANGELICAL-UNIONIST community of Szubin, 286 units — and “ydow” returns NOTHING. Szubin's pre-1874 coverage in this portal is Protestant and nothing else. SO THE ROUTE IS CLOSED HERE, though not necessarily in the world: the Szubin Jewish registers may survive unscanned at Bydgoszcz, at the Jewish Historical Institute in Warsaw, or in JRI-Poland, and NONE HAS BEEN ASKED. Bydgoszcz answered one letter in eight days, holds USC Szubin Miasto, and is the obvious place to ask. See szubin-pre1874-jewish-records.tsv. SEARCHED 27 SEPTEMBER, AND THIS ROW OVERSTATED THE OPPORTUNITY. The 119 scans are real and the figure reproduces exactly ("Found 119 scans from 34 teams, 104 units"), BUT THEY ARE NOT FRIEDERICA'S TOWN. Filtered to SZUBIN the same search gives TWO SCANS. The mass is WĄBRZEŹNO — 43, an Urząd Stanu Cywilnego of its own, 60 of the scans at the Toruń archive rather than Bydgoszcz — a hundred kilometres north, and probably another family entirely. The lane is narrow, not wide. AND IT STILL PAID, BECAUSE THE 1874 INDEX NAMES SIX PEOPLE AND THIS ARCHIVE READ FOUR: Sally Leibholz, Jacob Leibholz, Philipine Pollnow, Friederica Schmul, JUDA SCHMUL and JETTE MACHOT. HER PARENTS, NEVER HELD HERE. A Prussian marriage entry names the parents of BOTH parties, which this archive keeps saying about other people's records and did not apply to the one in its own hands. MACHOT has never been searched anywhere. AND THE INDEX SAYS PHILIPINE POLLNOW WHERE THIS ARCHIVE PUBLISHED WILHELMINE. Not the same name, both readings of one Kurrent hand, neither authoritative — she is on the front page and in the register as Wilhelmine and that is now PROVISIONAL until the scan is re-read with the alternative in mind. PLUS A CANDIDATE BROTHER: ABRAHAM SCHMUL, born Szubin 1846, died Koronowo 1904, father Juliusz, mother JETTE née MUSCHLL — same town, right generation, and a mother's forename matching Friederica's. Muschll against Machot is one name read twice; Juliusz against Juda is not. Both settling documents are pre-1874 Jewish congregational registers, which is exactly what row 266's Bydgoszcz letter asks about. Working in data/schmul-szubin-index-2026-09-27.tsv.

DRAFTED 22 September 2026, NOT SENT. requests/36-SEND-THIS-to-bydgoszcz-jewish-registers.txt, Polish with an English translation beneath it. Send as a REPLY on the thread of IU.631.434.2026 so it lands against the same file. IT IS NOT A REPEAT, ALTHOUGH IT LOOKS LIKE ONE. The letter of 15 September already asked whether Bydgoszcz holds the Szubin Jewish community records — but only as a conditional, “if the entry is not there after all”. The entry WAS there, so the condition never triggered and the question was never answered. The draft says so in as many words, so nobody at the archive thinks they are being asked twice. WHAT IT ASKS: whether the archive holds the akta gminy wyznaniowej żydowskiej w Szubinie — Jewish community registers from before 1874 — and if not, where they might now be or whether they were destroyed. FRIEDERICA SCHMUL was born at Szubin about 1852, and Prussian civil registration did not begin until 1 October 1874, so no Standesamt record of her birth can exist. IT OPENS BY THANKING THEM AND SAYING WHAT THEIR ANSWER PRODUCED — both documents, with signatures and scan numbers. That is deliberate: they pointed at a portal and two records fell out of it within the hour, and an archivist who hears that reads the next question more carefully. IT ALSO STATES THAT THE PORTAL HAS ALREADY BEEN CHECKED and names the one religious fonds Szubin does have, the Evangelical-Unionist community, so the archive is not asked to re-point at something already searched. TWO SUBORDINATE ASKS, both explicitly marked as less important: whether any finding aid would give SALLY LEIBHOLZ'S death date (the death registers have no skorowidze and the letter says it is not asking for a search), and — if the scan happens to be to hand — a glance at the word after “Tochter der” in akt 9/1874, which decides whether Friederica was the daughter of an unmarried woman or of a married couple. THIS ARCHIVE DRAFTS LETTERS AND DOES NOT SEND THEM. Nothing goes without David. SENT 22 SEPTEMBER 2026 on the IU.631.434.2026 thread, message 1a0c8d2a92207d0f. Text preserved at requests/sent/2026-09-22-bydgoszcz-jewish-registers-SENT.txt. ANSWERED 25 SEPTEMBER 2026, AND THE ANSWER CAME WITH A BONUS. Bydgoszcz replied under IU.631.434.2026. The formal letter is attached to the covering e-mail and NOT YET READ; what has been read is what Sandra Dończyk shared alongside it — USC SZUBIN MIASTO SYGN. 246, the alphabetical index to the Szubin death register 1894-1908, a volume this archive had published as non-existent. See data/szubin-sterbe-index-1894-1908.tsv.

RAISED AND WITHDRAWN THE SAME HOUR, 22 September 2026. The row said the Rosebank reply turned the SANDF service-record request from a guess into a document, by giving Leonard's occupation as ARMY and his residence as ROBERTS HEIGHTS. IT DID NOT, BECAUSE THE ARCHIVE ALREADY HAD BOTH — reconstructed from FamilySearch film 008161359 frame 475 on 17 September, published on the DNA page, and already used there to argue that he was a serving soldier. Request 16 was drafted with that evidence in it. NOTHING ABOUT THE SANDF ENQUIRY CHANGES. Struck rather than deleted, because raising it is the sixth time in one day this archive has treated something it already held as a discovery, and the count is the useful part.

DRAFTED 22 September 2026, NOT SENT. requests/37, to go as a REPLY on the existing Rosebank thread. THE THANKS COMES FIRST AND IS GENUINE: they answered in a week, courteously, with a scan attached, under no obligation to do either. IT DOES NOT TELL THEM THEY SENT THE WRONG BOOK. It was an easy mistake — the priest was also the State's marriage officer, so the parish holds the Government form B.M.D. 3 as well as its own register, and the difference is invisible from outside a presbytery. The letter simply NAMES THE TWO BOOKS IT WANTS — the LIBER MATRIMONIORUM and the BANNS REGISTER — says what is in them that the form lacks, and stops. AND IT DOES NOT MENTION that this archive already held every field on the scan they sent. That is true, it is recorded here, and saying it to them would be graceless and would make the next request less likely to be answered. It states the reason plainly — an elderly man in Australia who has never known who his father was, and a chain of evidence every step of which is circumstantial — and it offers the parish an easy exit: if the 1940 books are with the Archdiocese rather than the parish, just say so and the letter goes there instead. It discloses that David is not a descendant and makes no claim to be. REWRITTEN AND SENT 22 SEPTEMBER 2026, message 1a0c8d21508292fb. The draft as first written asked for the Liber Matrimoniorum and the banns register — TWO BOOKS ARLENE HAD ALREADY RULED OUT IN THE BODY OF HER EMAIL, which had not been read because only the PDF attachment had been opened. Caught while checking the thread for the right reply address. WHAT WENT IS A PURE THANK-YOU with no request attached, naming the three things she settled — the parish, the survival of the register, and the No on parents.

Derrick supplied his mother's own birth registration on 27 September and it settles what this archive had wrong. IVONE MARRIED FREDERICO PINTO DA ROCHA ON 18 NOVEMBER 1939 at the 1.ª Conservatória do Registo Civil de Lourenço Marques, and that marriage was dissolved by divorce, sentence of 6 DECEMBER 1945. Both are averbamentos on the assento. SO HENRIQUE, BORN 11 MARCH 1941, WAS BORN INSIDE THE MARRIAGE — sixteen months after the wedding and four years before the divorce. This archive had published that he was born to an unmarried mother and that his father 'gave him a surname and, on the evidence of that surname, acknowledged him'. The plainer reading is that his parents were married to each other. WHAT TO GET, IN ORDER OF VALUE: (1) THE MARRIAGE ENTRY OF 18 NOVEMBER 1939 — a Portuguese marriage assento names the parents of BOTH parties, so it should give Frederico's father and mother outright, and it is the single cheapest route into a family this archive has never touched. (2) Henrique's birth registration at Lourenço Marques, 11 March 1941, which the page already called the only document likely to name his father and which is now the second-best rather than the only. (3) The assento de óbito at Vila Nova de Gaia, which fixes the day of Ivone's death that the certificate's boxed digits will not give up, and which should now carry a death averbamento back onto the birth entry. WHERE: the Mozambique registers for this period are at the Arquivo Histórico de Moçambique, and the Conservatória dos Registos Centrais in Lisbon holds the transcriptions — it produced this very assento in 1987 under processo 10.277/87, which is a live reference to quote. AND THE CAUTION: a husband at the date of birth is a legal presumption, not a certificate. Nothing here excludes another father, and the archive should not swap one unevidenced paternity story for another.

Her birth assento gives her father's parish as VIMIEIRO and her mother's as NEGOZELA. BOTH ARE FREGUESIAS OF SANTA COMBA DÃO, DISTRITO DE VISEU — and Nagozela and TREIXEDO were merged into one freguesia in 2013, which means they adjoin. THE DNA PAGE'S LOOSE THREAD IS A TREIXEDO MATCH. Her mother came from the parish that match's parish was merged with. Every caution stands — the matches are distant and the closest Portuguese match still says only 'Portugal' — but the coincidence shrank from a district of half a million to one concelho of eleven thousand, reached from a register that knows nothing about DNA. AND TWO PEOPLE ARE ALREADY CATALOGUED AT THE TORRE DO TOMBO: JOAQUIM JULIO DE FIGUEIREDO, PT/TT/DGAP-DIA/001/056204, Processos de Funcionários do Ultramar, cx 6377 proc 103541, DATED 1922-1954 — an overseas civil servant's personnel file opening the year Ivone was born in the colony; and FREDERICO PINTO DA ROCHA, PT/TT/DGAP-DIA/001/059068, cx 6629 proc 106864, 1940-1983, plus a second file DGAP-DIA/002/7338. He married her in November 1939 and a service file opens the next year. SUCH A FILE CARRIES BIRTH DATE AND PLACE, PARENTS, MARITAL STATUS AND DEPENDANTS. NOT CONFIRMED: neither catalogue entry gives a parish, a date of birth or a parent, and the dates are the whole of the argument. CANDIDATES TO ORDER AND READ, not identifications. WHAT IS NOT AVAILABLE, and it is the finding rather than a failure to look: SANTA COMBA DÃO'S PARISH REGISTERS ARE NOT ON DIGITARQ. Viseu's holdings there are judicial and civil-government only, so the grandparents — António Júlio de Figueiredo and Ana Pereira de Figueiredo, Augusto Mata and Júlia Durães Mata — are out of reach online. They are at the Arquivo Distrital de Viseu and on FamilySearch microfilm. WHAT IS AVAILABLE NOW AND FREE: the Governo Civil de Viseu PASSPORT APPLICATIONS, PT/ADVIS/AC/GCVIS/H-D/001/xxxxx, digitised and indexed by individual name — which is the emigration itself, since the parents left for Lourenço Marques before March 1922. Also the Santa Comba Dão court records, PT/ADVIS/JUD/TJCSCD, whose inventários por óbito name heirs. AND THE TRAP IS THE LUTHERSTRASSE TRAP AGAIN: THERE ARE TWO VIMIEIROS, one in Santa Comba Dão and one in Arraiolos, Évora — and it is the ÉVORA one whose parish registers are digitised, so a name search returns the wrong parish first every time. The mother's Nagozela is unambiguous and settles which concelho; the father's Vimieiro is a reasoned preference until his own record says so. Full working in data/ivone-portuguese-origins.tsv.

CLOSED 27 SEPTEMBER, WITHIN THE HOUR, AND THE EXEMPLAR WAS SOMEWHERE THE ARCHIVE HAD ALREADY BEEN. This row was filed against the archive's own finding: the given name of Sally Leibholz's mother had been re-read as beginning with H — the clerk's capital P in POLLNOW is a round bowl and the disputed capital has none, so NOT the index's PHILIPINE; his capital H in HAMMERSTEIN has a flat baseline lead-in, two steep diagonals and A LOW CROSSBAR and the disputed capital has all three, so NOT this archive's WILHELMINE. THE STATED WEAKNESS WAS THAT KURRENT W AND H ARE ALIKE AND THERE WAS NO W IN THIS HAND ON EITHER PAGE OF THE SPREAD, so W was excluded on a positive H match rather than on a W exemplar. THE W WAS ON THE PREVIOUS SCAN, AND THE ARCHIVE ALREADY KNEW THE NAME. AKT NR 8 OF THE SAME VOLUME MARRIES HEINRICH GUSTAV JULIUS WENGATZ — the very near-miss the reader landed on in September while hunting for akt nr 9, recorded in leibholz-schmul-marriage-1874.tsv as "the right date and the wrong people". HIS CAPITAL W IS TWO ROUNDED LOOPING ARCHES, no straight strokes and no crossbar; the disputed capital is two steep straight diagonals joined low down. NOT THE SAME LETTER AND NOT CLOSE. The H reading now rests on a positive match AND a genuine W to fail against, and the page no longer carries the caveat. AND THE SAME SCAN GAVE A SECOND CONTROL NOBODY ASKED FOR: AKT NR 8 IS ALSO DATED DECEMBER, in the same hand, written heavily, same construction as akt nr 9's. That orders the volume — nr 7 on 24 November, nr 8 in December, nr 9 on 24 December, nine marriages from the day the office opened to the end of the year. THE NOVEMBER READING HAD NR 7 AND NR 9 SHARING A DATE WITH NR 8 STRANDED BETWEEN THEM. THE LESSON IS THE CHEAP ONE: the control was one click away in a unit this archive had already opened, described in its own notes, for five days. Crops published as pollnow-1874-control-W.jpg and akt8-december-1874.jpg. ROUTE, BECAUSE IT IS NOT OBVIOUS: a scan's modal id only resolves inside the current result set — open the modal, click "View the document" to expose the unit's scan list, click Show, and harvest the new Loris id from the network log. See pollnow-1874-reread-2026-09-27.tsv.

THE LETTER OF 22 SEPTEMBER TOLD A STATE ARCHIVE TWO THINGS THAT ARE NOW WITHDRAWN, AND THEY ARE SITTING IN ITS CASE FILE. It gave the marriage as "24 listopada 1874" — the register says DECEMBER — and named the groom's mother as "WILHELMINY z domu POLLNOW", a reading since withdrawn entirely on the clerk's own letterforms. IF ANYONE AT BYDGOSZCZ SEARCHED ON EITHER, THEY SEARCHED FOR SOMETHING THAT DOES NOT EXIST. The sent letter is deliberately left unedited, because it is the record of what was sent; the correction goes on the thread instead. AND THE LETTER OPENS BY ADMITTING THE LARGER ERROR, WHICH IS THIS ARCHIVE'S AND NOT THEIRS: it published that no alphabetical death index exists for Szubin, on the strength of a portal facet returning zero, and they answered by sending one. A CATALOGUE IS NOT AN ARCHIVE, and saying so plainly is both honest and the best way to make the next question worth an archivist's time. THE NEW QUESTION IS SMALL AND FOLLOWS FROM THEIR OWN GIFT: sygn. 246 covers 1894-1908. Sally Leibholz was alive in August 1884 and would have been only 46 in 1894, SO 1884-1893 IS THE LIKELIER WINDOW AND IS NOT IN THAT BOOK. Is there an equivalent Alphabetisches Verzeichnis for 1874-1893? The letter also tells them the L slot and the S/Sch slot have already been read in full, so that nobody repeats work their own parcel made possible, and re-raises the unanswered 22 September question about the pre-1874 Jewish community books. Drafted at requests/37-SEND-THIS-to-bydgoszcz-correction.txt. I DRAFT; SENDING IS DAVID'S. SENT 27 SEPTEMBER 2026, message 1a0e23f9d737321a, to kancelaria@ with sdonczyk@ in copy, on the IU.631.434.2026 thread. AND IT WAS SENT BEFORE THE ARCHIVE'S OWN ATTACHED LETTER HAD BEEN OPENED, WHICH ALREADY ANSWERED THE QUESTION IT ASKS. The reply of 25 September came as TWO e-mails — a chancellery note WITH THE FORMAL LETTER ATTACHED, and a separate folder share from Sandra Dończyk carrying sygn. 246. THE FOLDER WAS OPENED AND THE ATTACHMENT WAS NOT. The formal letter states that death indexes survive for USC Szubin Miasto 'wyłącznie z lat 1874-1878, 1894-1924' — so 1874-1878 EXISTS and 1879-1893 DOES NOT, and the letter that went out asks whether anything survives for 1874-1893. The corrections in it were right and worth sending; the question was asked of somebody who had already answered it in writing. THIRD TIME IN A WEEK for the same fault — read the new artefact, ignore the context around it. WHAT TO ASK NEXT, AND IT IS NOW SHARPER THAN WHAT WAS SENT: (a) the 1874-1878 death index, which exists and is unread; (b) THE JEWISH COMMUNITY REGISTERS 1820-1847, because ABRAHAM SCHMUL WAS BORN AT SZUBIN IN 1846 and his entry would name his parents in a registrar's hand — the document that decides whether he is Friederica's brother. See data/bydgoszcz-reply-2026-09-25.tsv.

BYDGOSZCZ CONFIRMED ON 25 SEPTEMBER THAT IT HOLDS THE AKTA METRYKALNE ŻYDOWSKIEJ GMINY WYZNANIOWEJ W SZUBINIE, 'JEDYNIE Z LAT 1820-1847'. The question was asked for FRIEDERICA SCHMUL, born about 1852-1854, AND SHE IS NOT IN THEM — the answer to the question as put is no. BUT ABRAHAM SCHMUL WAS BORN AT SZUBIN IN 1846, INSIDE THE SPAN BY ONE YEAR. He is the candidate brother of schmul-szubin-index-2026-09-27.tsv, held only on a volunteer's indexed card giving father JULIUSZ and mother JETTE née MUSCHLL, against the 1874 marriage index's JUDA SCHMUL and JETTE MACHOT and the Bydgoszcz archivist's JUDA and JUTTA née MACHOL. THREE READINGS OF ONE COUPLE AND NOT ONE OF THEM FROM A REGISTER. HIS 1846 BIRTH ENTRY WOULD NAME HIS PARENTS IN A REGISTRAR'S HAND and settle both questions at once: whether he is her brother, and what the mother's maiden name actually is. A MARRIAGE of Juda and Jutta, if they married in the 1840s, would also fall inside the span and would be better still. ASK FOR: the 1820-1847 Szubin Jewish registers — whether they are digitised, whether they can be shared as sygn. 246 was, or what a query costs. Bydgoszcz has now answered three letters and shared a 76 MB volume unprompted; this is a reasonable next ask and should be made warmly and once. 11 OCTOBER — DRAFTED, AND THE PRE-CHECK THAT WOULD HAVE MADE IT UNNECESSARY COULD NOT BE RUN. The letter is at requests/40-SEND-THIS-to-bydgoszcz-1820-1847.txt, Polish with an English translation, to go as a reply on the IU.631.434.2026 thread. IT ASKS ONE QUESTION: Abraham Schmul's 1846 birth record, or what a one-year query costs — a single year inside a twenty-seven-year span, not a search. It opens by telling them what their last answer produced, INCLUDING THAT THEIR ARCHIVIST'S READING CORRECTED THIS ARCHIVE, because an archivist is rarely told that. AND IT ADMITS SOMETHING RATHER THAN DRESSING IT UP. The September letter's best sentence was “I have checked the portal so as not to waste your time”, and that check cannot be repeated today: on 11 October GENEALOGIAWARCHIWACH.PL WAS DISPLAYING ITS OWN NOTICE of difficulties while work proceeds on reconfiguring its internet links, its default search runs in SCAN MODE over indexed names rather than over the catalogue (“Szubin żydowskiej” returns zero, “Szubin” returns two scans from two units, neither a Szubin fonds), and its fonds browser would not open. AND SZUKAJWARCHIWACH.GOV.PL, THE NATIONAL CATALOGUE THAT WOULD ANSWER THE QUESTION DIRECTLY, IS BEHIND AN INCAPSULA BOT GATE — RECORDED AS BLOCKED AND NOT WORKED AROUND. The letter says all of that plainly instead of pretending the check was made. I DRAFT; I DO NOT SEND. This one is David's to read and press.

David obtained a Trove API key on 27 September 2026, VALID TO DECEMBER 2026. The key is in notes/trove-key.txt and notes/ is gitignored; tools/trove.py reads it from there or from TROVE_API_KEY. IT MUST NEVER BE COMMITTED. THE COVERAGE FACT THAT MAKES IT USEFUL: Australian newspaper digitisation stops at 1954 for copyright, so Trove is normally useless for a family that arrived in 1965. TWO EXCEPTIONS AND BOTH ARE OURS — THE CANBERRA TIMES IS DIGITISED 1926-1995 and the COMMONWEALTH OF AUSTRALIA GAZETTE 1901-1973. The family landed in Canberra. Control: "Canberra" returns 5,029,116. FOUND: Joachim Peter Leibholz gazetted as naturalised, Commonwealth of Australia Gazette No. 193 of 28 September 1944 p.2212, "Leibholz, Joachim Peter, [6]7 Walsh-street, South Yarra, Victoria". AND IT IS SMALLER THAN IT LOOKED: australia.astro has held the naturalisation since 21 September (granted 27 July 1944, South Yarra, previously STATELESS) and his arrival to the day (Anchises, 12 June 1939). NEW IS TWO THINGS ONLY — THE STREET, where this archive had the suburb, and a second independently produced public attestation of a fact that rested on one NAA card. AND THE CANBERRA SEAM IS REAL AND MOSTLY UNPUBLISHABLE: nine "Luwinski" hits, EIGHT of them Canberra Times 1970-1987, exactly where the 1965 roll and the Marquardt sponsorship put the family — and they name the children, who are living. Recorded in notes/2026-09-27-trove-canberra-PRIVATE.md and NOT on the site. Publishable: the family were in Canberra by 1972, and a January 1973 family notice attests MANFRED AND HELGA BOTH ALIVE AND IN CANBERRA. WARNING THAT CUTS AGAINST THIS ARCHIVE'S OWN METHOD: TROVE NORMALISES SPELLING. "Luwinski" and "Luwinsky" return the IDENTICAL nine articles. A Trove search is NOT a spelling test and cannot be used as one. NEXT, IN ORDER: (1) the Commonwealth Gazette naturalisation lists for MARQUARDT, which row 215 says would settle the sponsor question because WHAT DECIDES IT IS AN ADDRESS — a 1967 list is already in hand unread; (2) the Australian Jewish News, Melbourne, 1969 WIZO item and the 1952-55 Melbourne social columns, for the Leibholz lane; (3) shipping notices for the FLAVIA, Bremerhaven to Australia, arriving 17 September 1965. MANFRED RETURNS ZERO — he died at Brisbane in 2012, past the 1995 cut-off, so row 213's Ryerson Index is still the only free instrument that reaches his death. WORKED THROUGH AND CLOSED 27 SEPTEMBER. THE BEST ITEM WAS NOT THE GAZETTE. THE ARGUS, 18 February 1944, carries JOACHIM PETER LEIBHOLZ'S OWN NOTICE OF INTENTION: "of German nationality, born at Berlin, resident 4 YEARS 8 MONTHS in Australia, now residing at 67 WALSH STREET, SOUTH YARRA". IT FIXES THE HOUSE NUMBER the gazette OCR lost (67, not "07") AND CORROBORATES THE ANCHISES TO WITHIN DAYS — 4 years 8 months before February 1944 is June 1939, and the passenger record says 12 June 1939. A man counting his own months against a shipping list compiled by somebody else. AND A DISCREPANCY THAT IS NOT A CONFLICT: he declares himself GERMAN where the NAA card says STATELESS. The Eleventh Decree of November 1941 made him stateless; he wrote down what he was born. HE WENT BY PETER — The Age, 2 October 1954, "Mr. and Mrs. Peter Leibholz" at a Toorak ball among Nelken, Horowitz, Abry. Searching JOACHIM finds the gazette; searching PETER finds the man. MRS. H. LEIBHOLZ joins a Melbourne WIZO group, Australian Jewish News 8 August 1969 — AN INITIAL OF H against the archive's own "Tilde on one file, Hilde on another". Supports one reading, proves neither. A DAUGHTER, Dr Jane Leibholz, m. Stewart Wallace 1968 (Australian Women's Weekly) — MAY BE LIVING, NOT IN THIS TREE, NOT PURSUED. AND THE SPONSOR QUESTION MOVED: MARQUARDT, ANNE MARIE, 2 KENNY PLACE, HACKETT, naturalised 28.9.66 — a Marquardt household at Hackett with a street and a number, which is what row 215 said would decide it. H. MARQUARDT IS NOT IN THE GAZETTE LISTS AND THAT WAS CHECKED: five further lists read, every Marquardt in them is North Ryde, West Hoxton, Longreach or Montrose, none in the ACT. If he came on the ANNA SALEN in 1955 he was eligible around 1960, before these lists. THE ABSENCE IS EXPLAINED. THE FLAVIA PAID NOTHING: Trove's shipping coverage dies with the newspapers in 1954 and the twelve hits are 1964 Dutch Australian Weekly advertisements for other voyages. TROVE IS NOW A KNOWN INSTRUMENT AND THE KEY RUNS TO DECEMBER 2026. What is left in it for this family: the Canberra Times 1970-1987 seam, which is living people and stays in notes/.

RAISED BY A PEER SESSION ON 27 SEPTEMBER AND VERIFIED HERE BEFORE ANYTHING WAS CHANGED. NARSSA'S OWN SEARCH PAGE SAYS IT: "NARSSA is in the process of migrating more than 8.3 MILLION ENTRIES from the old NAAIRS database to the new database and therefore 'fulltext search' (new database) WILL NOT DISPLAY ALL THE RECORDS IN THE SYSTEM. In the meantime NARSSA makes both databases available for use by the public. Using the 'Search Old NAAIRS database' WILL PROVIDE A MORE COMPLETE SEARCH." Read at nationalarchives.gov.za/node/737, 27 September 2026. WHY IT MATTERS HERE: every NAAIRS bound this archive wrote is about THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NOT ONE IS ABOUT THE DATABASE'S COVERAGE OF THE INDEX, because there was no reason on 15-16 September to think a database could be partial. EVIDENCE THAT THE OLD COMPLETE ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: the searches use the old database's vocabulary — 'all repositories', 'national registers', 'central-government database', the code RSAE — which the new fulltext search does not have; and the 16 September rows name the sm300 Search Manager endpoint, which IS the old one. NOT ESTABLISHED FOR THE 15 SEPTEMBER ROWS: row 140 records only that the pre-move entry point 'stopped resolving' and never wrote the URL down. A record-keeping failure, and the reason the bound was needed. SIX NEGATIVES BOUNDED, NOT WITHDRAWN. The peer listed eight; TWO DO NOT BELONG — Ivone's naturalisation-notice nil was searched on gazettes.africa and never touched NAAIRS, and the 'NAAIRS is down' row is already a withdrawal with no live claim. BLOCKED, AND BY THE SAME THING AS ROW 178: probed 27 September, www.national.archsrch.gov.za/sm300cv/smws/sm300dl answers 503 and naairs.htm is 404. THE OLD DATABASE IS DOWN AND THE NEW ONE IS BY NARSSA'S OWN STATEMENT INCOMPLETE, so there is currently NO INSTRUMENT THAT CAN CLEAR THESE NILS. WHEN IT ANSWERS: re-run surname WEAR estates 1970-76, IMMIGRATION+LUWINSKI, PIO+LUWINSKI, NATURALIZATION, LONGHURST, LEIBHOLZ, and the four variant spellings — ON THE OLD DATABASE EXPLICITLY, RECORDING THE URL. THE RULE EARNED, AND IT IS GENERAL: an instrument can be partial without saying so at the point of search. This archive asks whether an index covers the records; IT HAD NOT THOUGHT TO ASK WHETHER THE DATABASE COVERS THE INDEX. Record the endpoint with every nil, not just the query. See data/naairs-two-databases.tsv. || 27 SEPTEMBER, LATER: A CORROBORATION FROM THIS ARCHIVE'S OWN NOTES, AND A TENSION RECONCILED. The gazettes.africa row — the one that is NOT a NAAIRS negative — carries in its bound, written 16 September and for an entirely different purpose: "...which are indexed on NAAIRS and WERE NOT SEARCHED BECAUSE NAAIRS WAS UNREACHABLE ON THE DAY." CONTEMPORANEOUS DATED EVIDENCE OF THE OUTAGE, from before anyone knew a migration was behind it. AND IT SITS AGAINST ROW 22, which says four searches ran that same day "on the corrected endpoint". Those cannot both be true of one homogeneous day — and row 140 reconciles them in order: the old entry point STOPPED RESOLVING, the service was found to have MOVED to the sm300 address, and the searches were run there. THE DAY HAS A BEFORE AND AN AFTER. THIS STRENGTHENS THE IDENTIFICATION WITHOUT CLEARING ANYTHING: it puts the 16 September nils on sm300 — the old complete database — by this archive's own timeline rather than only by their vocabulary, and it makes the 15 September nils likelier to be the same, because row 140 describes A SERVICE THAT MOVED, NOT A DATABASE THAT CHANGED, and a URL move does not swap a corpus. THE BOUND STANDS UNCHANGED. The pre-move URL was never recorded, and a nil is not cleared on a reconstruction however tidy. GENERAL LESSON: A DATED NIL IS NOT ATOMIC. '16 September' contained an outage, a discovery and a migration. When auditing searches against an instrument that moved, look for a change WITHIN the day. 11 OCTOBER — THE BLOCKER IS GONE AND THE ENDPOINT NOW HAS A NAME. NAAIRS returned at https://www.nationalarchives.gov.za/search/, the “NATIONAL ARCHIVAL INFORMATION SEARCH SYSTEM” — free, no account, all twenty repositories, with a JSON API at /search-api/api/search. The old Search Manager endpoint is still 503 and naairs.htm still 404, so the service came back somewhere else for the second time in a month. AND THE TWO-DATABASE QUESTION THIS ROW EXISTS FOR IS ANSWERED FROM NARSSA'S OWN PAGE: the migration warning is still published — 8.3 million entries, the “fulltext search” admittedly incomplete — and it ends “To search the complete database please click on: Search NAAIRS Database”, WHOSE LINK IS THAT PORTAL. So the “old complete database” this row went looking for as a separate address IS THIS ONE. Three of the six nils were re-run the same day under seventeen spellings with SMITH (47,936) and SCHMIDT (3,418) as live controls, AND ALL THREE HELD. In data/naairs-is-back-2026-10-11.tsv. AND THE SIX NILS ARE RE-RUN. Three on the morning of 11 October under seventeen spellings (see data/naairs-is-back-2026-10-11.tsv) and the rest the same afternoon with MOTA, SALKOW, RAPPOPORT and FIGUEIREDO read in full. ALL HELD. The endpoint is recorded, the database is the one NARSSA itself calls complete, and A REFERENCE CODE CANNOT BE SEARCHED ON IT — so a citation is re-verified by surname or not at all.

Drafted 27 September after the tree page was rebuilt to carry BOTH LINES — the blood line and the line of the name. THE LARGEST HOLE IN THE ARCHIVE IS NOT IN ANY REGISTER. The record says Kurt married Ivone in June 1948 and filed for divorce the same year, WITHIN MONTHS OF THE WEDDING AND BEFORE THE CHILD WAS BORN — and that he died in JOHANNESBURG ON 15 DECEMBER 1968, the same city, for twenty years. The archive therefore claims neither that he raised Derrick nor that he did not, because nothing has been read either way. ONE CONVERSATION CLOSES IT. AND THE SECOND SECTION IS THE ONE THIS ARCHIVE DID NOT EXPECT TO BE ABLE TO ASK: Kurt's family did not all die in the war. AUGUSTE 'LIZZY' LUWINSKI, his mother, died at Kraichtal 22 JUNE 1975. SIEGFRIED LEIBHOLZ, his brother, hidden near Sommerfeld from January 1943 to April 1945, died 1 FEBRUARY 2005. Pauline died 1985. THERE WERE DECADES IN WHICH CONTACT WAS POSSIBLE and the archive knows nothing about whether there was any. Thirty questions in seven sections: Kurt; the Berlin family who survived; what was known in the family and when; Ivone and Portugal, including whether she wrote MOTA or MATA herself; Henrique and whether more of the 16 March 2010 fax survives — ONE SHEET OF IT MAY BE THE 1939 MARRIAGE ENTRY still being chased; things rather than memories; and a final section marked SKIP FREELY, including when he learned Kurt was not his biological father and whether anything on the site should come down. HOW ANSWERS WILL BE HANDLED, AND IT IS SAID IN THE DOCUMENT: family testimony, attributed and dated, A DIFFERENT GRADE OF EVIDENCE FROM A DOCUMENT, recorded as such and never silently blended with the records. A 'don't know' is a real answer and gets recorded as one. CHECKED BEFORE SENDING: the sheet carries no birth year, age or age-derived tell for Derrick. I DRAFT; PUTTING IT TO HIM IS DAVID'S. || ANSWERED IN PART, 28 SEPTEMBER 2026, WITHIN A DAY OF BEING WRITTEN. David: "KURT RAISED DERRICK." THAT CLOSES THE LARGEST HOLE IN THE ARCHIVE. Kurt is not the biological father — the DNA exclusion is untouched — AND HE IS THE MAN WHO BROUGHT HIM UP. Both now published, on the tree, on Derrick's page and in data/family-testimony.tsv, graded FAMILY and attributed. AND IT EXPOSED A DOCUMENT THIS ARCHIVE HAD BEEN READING WITH A LEAN: WLD 1950/1948 was an action for RESTITUTION OF CONJUGAL RIGHTS, granted only in 1951 — in South African practice the first required step of a divorce, not the end of a marriage. So the marriage ran on paper about three years and the child was born inside it. THE ARCHIVE HAD WRITTEN 'within months of the wedding and before the child was born', which is true and which leaned. A FILING DATE IS NOT A DEPARTURE DATE. WHAT IS NOW THE MOST VALUABLE PART OF THE SHEET: QUESTIONS 7 TO 10. If Kurt raised him, then AUGUSTE 'LIZZY' LUWINSKI — who survived the war and died at Kraichtal 22 June 1975 — was the grandmother of the house he grew up in, and SIEGFRIED LEIBHOLZ, hidden near Sommerfeld 1943-45 and alive until 1 February 2005, was his uncle in fact. Whether there was contact between Johannesburg and Germany is now a question about his own childhood rather than about a chart. STILL OPEN from section 1: for how long, where, whether before or after the 1951 decree, and whether he grew up in Kurt's house or Ivone's.

NEW 29 SEPTEMBER 2026, FROM THE LANDESARCHIV'S OWN LETTER. The Sammelakte to the Schöneberg II marriage register EXISTS and holds “nur die Aufgebotsunterlagen aus dem Jahre 1923” — the banns papers. A German banns file of 1923 normally contains THE BIRTH CERTIFICATES OF BOTH PARTIES, and a birth certificate names the parents. THE 1924 MARRIAGE ENTRY NAMED NO PARENTS — the 1920s Prussian form cites the birth register instead — so JOACHIM LUWINSKI, JOHANNA née THIMM, SALLY LEIBHOLZ and FRIEDERIKE née SCHMUL are still unconfirmed on one document, and this box is the nearest thing to one. NOT DIGITISED, so no scan; copies are ordered through the Digitaler Antrag at service.berlin.de/dienstleistung/352800/. THIS IS A PURCHASE AND IT IS DAVID'S TO MAKE. DRAFTED, NOT ORDERED. AND A CLEAN NEGATIVE SO NOBODY ORDERS IT FOR THE WRONG REASON: the Sammelakte holds NO DIVORCE PAPERS. The judgment is Landgericht Berlin 251. R. 80. 40 and it lives, if it lives, in a court series. SECOND CANDIDATE ON THE SAME LOGIC: Auguste's own birth entry, GROß-BÖSSAU 1893 NR. 88, which the marriage entry cites and which this archive had no way to ask for until today.

DRAFTED 29 SEPTEMBER 2026 as requests/39-REPLY-to-landesarchiv-oberle.txt. NOT SENT — I draft letters, I do not send them. IT DOES FOUR THINGS. (1) It supplies STANDESAMT BERLIN VIII, GEBURTSREGISTER 1912, NR. 963 with the reasoning, which is exactly what Frau Oberle wrote that she lacked, and says plainly that the failure was ours and not hers. (2) It offers her GERTRUD'S MAIDEN NAME, LEIBKE — her own register enters father, mother and marital status as not known, and this archive can supply the surname from an AJDC card that matches her entry on address and date of birth. A reply that only takes is worse than one that gives something back. (3) It confirms that the divorce note resolves the 1940 “Witwe” problem, so she knows what her scan was worth. (4) IT ASKS ONE QUESTION BEFORE ANY MONEY IS SPENT: do the 1923 Aufgebotsunterlagen customarily contain BOTH PARTIES' BIRTH CERTIFICATES? The parents are the entire reason to order the Sammelakte, and the marriage entry names none. If yes, order; if no, go at the two birth registers directly and save the fee. SEE ALSO ROW 229, the supplement drafted on 20 September and still unsent — this letter supersedes it and should go instead.

A peer session is moving the estate onto subdomains and sent a recipe. THE WORK IS DONE AND SITS UNMERGED ON claude/subdomain-step4. DO NOT MERGE IT WHILE THE PAGES cname IS null. Measured 12 October: gh api repos/daviddef/TheLuwinski/pages --jq .cname returns null, so the site is served at daviddef.github.io/TheLuwinski and base must stay '/TheLuwinski'. Pushing the branch now would make every link root-relative on a host that serves the site from a subdirectory — the exact breakage described below, self-inflicted. THE ORDER IS: David says go, the peer binds AT THE MOMENT OF THE PUSH, not before. WHAT THE BRANCH CONTAINS, AND THREE OF THE FIVE ARE NOT IN THE PEER'S RECIPE: site/public/CNAME; astro.config site and base; THE FOUR REDIRECT TARGETS (/letters pointed at /TheLuwinski/errands/ and would have 301'd into a 404); Base.astro's ORIGIN constant, which is the trap of being OUT of the kit ring — the siblings read their origin from archives.json via SocialHead and this archive is deliberately not in that file, so the origin is a local constant and a scratch build caught it still reading daviddef.github.io while base and every link were already right; and a link that was never a link, sources.json carrying <a href={'/TheLuwinski/negatives/'} inside a JSON string so it rendered as a literal href={'...'} and has been dead on the live site all along. THE LESSON, WHICH COST THREE WORKING SITES ACROSS THE ESTATE: the recipe said binding first costs nothing because GitHub keeps serving the previous build. It does — FROM A HOST WHERE THAT BUILD'S OWN PATHS DO NOT EXIST. Binding 301-redirects the project URL to the bound domain, so it moves every visitor onto a site whose base no longer matches. Luwinski, Falco and Mazza were all taken down this way (89, 308 and 205 dead hrefs) until measured and unbound. THE GAP IS NOT ELIMINATED BY REORDERING, IT IS ONLY MOVED: landing config before binding breaks the site and binding before config breaks it too, whenever base is not '/'. The only fix is to make the two simultaneous. TWO MEASURED FACTS TO CHECK AFTER ANY FUTURE BINDING: public/CNAME DOES NOT BIND a domain — the API can read cname null with the file sitting in the deployed output; it only stops a later deploy tearing an existing binding off. And https_enforced can be false with the certificate approved, leaving plain http serving 200 instead of redirecting, so every page has two addresses; OURS IS ALREADY true, measured. AND THE RING STAYS SEVEN: Luwinski is not in kit/data/archives.json, removed 16 September at David's instruction because the Luwinskis are not a branch of the same family, and the peer's step zero would have put it back. Case (b): this archive does not use the kit ring at all. DONE. David said “push”; the branch was merged at eb1e914 and a peer session bound the domain the moment the run completed. MEASURED TWICE, A MINUTE APART, AND BY THE CORRECTED METHOD — every root-relative link the LIVE PAGE EMITS, fetched verbatim, rather than paths constructed from what ought to be there. Home page: 40 distinct emitted links, ALL RESOLVE. /leibholz/: 49, ALL RESOLVE. The only non-200 in either sweep was //gc.zgo.at/count.js, which is PROTOCOL-RELATIVE AND EXTERNAL — an artefact of my own script prefixing it with our host, not a broken link; gc.zgo.at/count.js answers 200. https 200 with clean TLS on both passes, plain http 301s to https, cert APPROVED, https_enforced true. AND EVERY OLD ADDRESS 301s WITH ITS PATH PRESERVED — /TheLuwinski/tree/, /leibholz/, /errands/, /dna/ all land on the same page at the new origin, so no link anyone has ever shared is lost. The four redirect stubs follow to 200. THE TLS RISK WAS REAL AND UNTESTABLE AND DID NOT BITE. The plan was to lower https_enforced before binding and raise it after; GITHUB REFUSES — PUT returns 404 “The certificate does not exist yet”, so that lever does not exist and nobody should design around it. We went in with no way to de-risk it, said so, and set an abort condition the peer could act on without consulting us. On a REBIND of a previously bound hostname the certificate came back approved within seconds. NOT GENERALISED TO A FIRST BIND of a never-bound name: one data point, on a hostname that had held a cert hours earlier. AND THE RULE WE BOTH HAD WRONG, now settled by comparing two archives bound seconds apart: THE WINDOW IS NOT BIND-VERSUS-PUSH, IT IS THE DISTANCE BETWEEN THE BIND AND THE DEPLOY LANDING, whichever order they fall in. Ours paid nothing because the deploy had already completed when the bind landed. Falco was bound three seconds BEFORE its run started and was 404 for 87 seconds. So the trigger is BIND AFTER THE RUN COMPLETES, not when it leaves the queue.