What has actually been read

The fault this page was built to close 17 September 2026

This archive kept twenty-one evidence files in data/ — censuses, the Gedenkbuch, the 1939 Register, passenger lists, court files. Not one build script read any of them. The register was assembled from the MyHeritage export and corrections.json and nothing else. Everything a record actually said reached this site only because a human retyped it into a paragraph.

That is survivable while the human is paying attention and silently wrong the moment they are not. Nothing would have noticed if a file had been added and never surfaced, or if a page and a file had come to disagree. There was no gate, because there was nothing to gate.

The clearest casualty was Samuel Israel Leibholz. The register printed his birth as “June 28 1877?” — with the source tree’s question mark intact — while data/leibholz-gedenkbuch.tsv held the Bundesarchiv’s entry en909753, giving 28 June 1877 at Schubin, and his death at Auschwitz on 10 April 1943. The archive had the citation. The page showed the doubt.

What this page does not do

It does not rewrite the tree. The register and the charts draw the export as given, and that rule has not changed: the question mark on Samuel’s birth year is still there, because that is what the source says. What has changed is that the record sits beside it instead of in a file nobody reads. Where a named source genuinely overrides the family record that is a correction, shown as was and now, and it is a different mechanism on purpose.

Names this builder refuses to resolve Not linked

A name match is not a person match, and the first version of this builder proved it on its own data. The register holds a Kurt Luwinski who died in 1995 — a surname-index entry — and a Kurt Israel Luwinski, 1912 to 1968, who is the man this archive is about. Matching the court files on the bare name hung three divorce actions on the wrong one. The builder now pools every candidate before deciding, and where more than one person answers to a name it links neither and says so here.

Johannes Antonie Oelofse DE WITT — documented in litigation.tsv — could be: Johannes Antonie Oelofse De Witt #9500160, d. Apr 2 1963 | Johannes Antonie Oelofse de Witt #9500005, d. 2 April 1963, Sasolburg

KURT LUWINSKI — documented in arolsen-leibholz-luwinski.tsv, familysearch-sa-family.tsv, litigation.tsv — could be: Kurt Luwinski #9500040, d. 1995 | Kurt Israel Luwinski #9500001, d. 15 December 1968

Resolving these is a human job and is deliberately not automated. The second pair above is also a duplicate in the source tree that the register's own collapse list does not cover — one copy carries a death date and the other does not.

Four files are keyed by reference and not by person

adressbuch-boroughs-1938-1943.tsv, adressbuch-fulltext-sweep.tsv, adressbuch-namenteil-runs.tsv, agnes-leibholz-kerger.tsv, alfred-luwinski-sachsenhausen.tsv, arolsen-tracing-files.tsv, atlas-pin-audit.tsv, australia-indexes.tsv, berlin-adressbuch-names.tsv, berlin-adressbuch.tsv, berlin-directories-before-1903.tsv, berlin-marriage-indexes-are-blind.tsv, berlin-namensverzeichnisse.tsv, berlin-postwar-luwinski.tsv, bydgoszcz-reply-2026-09-25.tsv, caroline-james-1920-1957.tsv, castle-1911-meeanee-barracks.tsv, castle-baptisms-weston.tsv, castle-frederick-soldier.tsv, castle-giddings-1862.tsv, charlotte-leibholz-tempelhof-1982.tsv, chodowieckistrasse-1913.tsv, dewitt-confirmed.tsv, dieffenbachstrasse-58a.tsv, east-berlin-1990-1994-deaths.tsv, ella-leibholz-tiergarten-1974.tsv, entschaedigung-berlin.tsv, ermland-sauerbaum.tsv, family-testimony.tsv, familysearch-south-africa.tsv, findmypast-newspapers.tsv, gazette-wear-sweep.tsv, gazettes-africa-sweep.tsv, gedenkbuch-fifty.tsv, gedenkbuch-leibholz-findings.tsv, geneteka-pomorskie-2026-10-11.tsv, gerda-leibholz-marriage-chase.tsv, gertrud-leibholz-sterbeeintrag-1943.tsv, heinrich-leibholz-kreuzberg-1948.tsv, hilde-leibholz-wilmersdorf-1983.tsv, ivone-assento-nascimento-1922.tsv, ivone-death-certificate-2010.tsv, ivone-portuguese-origins.tsv, jacob-leibholz-deportation.tsv, joachim-peter-leibholz-birth-1903.tsv, klara-leipsholz-kreuzberg-1960.tsv, kreuzberg-1938-1979-deaths.tsv, kreuzberg-ocr-control.tsv, kreuzberg-reinickendorf-deaths.tsv, kurt-birth-entry.tsv, kurt-papers-1934-1939.tsv, labsa-berlin-catalogue.tsv, labsa-charlottenburg-inventory.tsv, landesarchiv-reply-2026-09-29.tsv, leibholz-luwinski-1924-entry.tsv, leibholz-luwinski-marriage.tsv, leibholz-registration-card.tsv, leibholz-shanghai.tsv, leipholz-postwar-berlin.tsv, leipholz-sweep.tsv, leipholz-telephone-slot-audit.tsv, lutherstrasse-34-standesamt.tsv, luwinski-emigration-1965.tsv, luwinski-germany-2001.tsv, melbourne-leibholz-death.tsv, naa-recordsearch.tsv, naairs-figueiredo.tsv, naairs-is-back-2026-10-11.tsv, naairs-sweep-2026-10-11b.tsv, naairs-two-databases.tsv, negatives-classified-2026-10-10.tsv, negatives-classified.tsv, negatives-control-audit.tsv, oskar-leibholz-theresienstadt.tsv, other-offices-deaths-2026-10-07.tsv, pollnow-1874-reread-2026-09-27.tsv, probate-calendar-wear.tsv, register-date-audit.tsv, reinickendorf-1982-1994-deaths.tsv, rosenberg-of-schlochau.tsv, schmul-and-thimm.tsv, schmul-half-of-the-1874-entry.tsv, schmul-szubin-index-2026-09-27.tsv, schoeneberg-1938-1968-instrument.tsv, schoeneberg-1939-guenther-leibholz.tsv, schoeneberg-death-indexes.tsv, schoeneberg-marriage-indexes.tsv, schoeneberg-progress.tsv, spandau-1980-1982-deaths.tsv, spandau-1982-leibholz-schoenfeld.tsv, surname-lines.tsv, szubin-civil-registers.tsv, szubin-death-registers-searched.tsv, szubin-pre1874-jewish-records.tsv, szubin-registers-online.tsv, szubin-sterbe-index-1894-1908.tsv, tna-discovery.tsv, tree-audit.tsv, treuchtlingerstrasse-2-4.tsv, trove-2026-09-27.tsv, trove-and-ryerson-22-september.tsv, twenty-alone.tsv, twenty-next.tsv, wear-1939-bt27-images.tsv, wear-bloom-1842.tsv, wear-castle-wedding-1914-newspapers.tsv, wear-newington-baptisms-1833.tsv, wear-pow-1943-45.tsv, weinstrasse-pages-of-testimony.tsv, west-berlin-deaths-2026-10-08.tsv, westpark-films.tsv, worklist-triage-2026-10-10.tsv, ydna-s11477-2026-10-10.tsv, ydna-s11477-enrichment.tsv — these record a search, a sweep or a set of archival references rather than a household. They are declared in the builder’s manifest so that they are accounted for, and they attach to no individual on purpose.

The gate that now stands behind this page

tools/check-evidence.py runs on every build. It refuses the build if a file appears in data/ that the builder does not declare, and it refuses the build if a person named in an evidence file reaches the published site carrying neither a record nor a correction. The failure that made this page possible cannot happen again without the build stopping.

Why this page exists

Several of this archive’s conclusions now rest on absences. That Leonard has no South African estate. That Ivone never naturalised. That the marriage registers never asked for parents. Each of those is only worth something because somebody measured how far the search reached — and until now those measurements lived in commit messages and working notes, where a later reader would never find them.

The distinction the page is trying to hold is between this is not there and I did not find it. The first is a finding. The second is a shrug. The column that separates them is the last one.

Kurt Luwinski's estate was never advertised in the South African Government Gazette — or the scanner lost it

Searched: gazettes.africa, the full South African run. LUWINSKI and LUWINSKY, every year; then the OCR manglings Luwinsk, Luwenski, Lnwinski, Luwiuski, Luwinshi; then Lewinski, whose two hits in 1969 and 1970 were opened.

Result: EIGHT documents in the entire corpus, and not one is his estate: three in October 1942 (16th, 23rd, 30th) dissolving the partnership between Walter Oppenheimer and Kurt Luwinski, THREE in November 1941 (7th, 14th, 21st) creating it, one in November 1972 for Paula Luwinski and one in July 1975 for Paula Luwinski at Pallou's Dress Centre, 18 Wanderers Street. *** THAT ENUMERATION WAS WRONG HERE UNTIL 22 SEPTEMBER 2026 AND THE ERROR WAS THIS PAGE'S ALONE: it said eight and then listed six, giving two November 1941 notices instead of three and omitting the 1975 one altogether. The Kurt page has carried all eight correctly throughout, and so has the work list. Re-counted against the live search, which still returns exactly eight. *** NOTHING AT ALL IN 1968, 1969, 1970 OR 1971. The two Lewinski hits are both VON LEWINSKY in a tariff list.

What would fall outside it: THE CONTROL IS WHAT MAKES THIS WORTH ANYTHING, and it kills the innocent explanation. The years are fully in the corpus and full of exactly the kind of notice that is missing: searching 1969 for 'Estates of deceased persons' returns 221 issues, 'boedel' returns 100, 'Master of the Supreme Court' returns 111; 1970 gives 163, 93 and 101. So this is not a gap in the archive of gazettes. WHAT IT STILL CANNOT DISTINGUISH: a notice the scanner mis-read from a notice that was never placed. The search does no prefix matching — 'Luwinsk' returns zero where 'Luwinski' returns eight — so a mangled spelling is invisible to it, and this corpus has already been caught rendering Figueiredo as 'Epo .Figueiredo,CarlosDe'. He died on 15 December 1968 and his estate is MHG 7584/69. Only the file itself can say which.

Leonard Aubrey Robert Wear is in no South African record FamilySearch holds except his two marriages

Searched: ALL THIRTY-TWO South African collections FamilySearch holds, found by enumerating all 3,502 of its collections and filtering. Surname WEAR matched EXACTLY - 111 records, every one read - then repeated under WARE, WEIR, WIER, WEARE, WEARNE and WEARS.

Result: He appears TWICE. Johannesburg, 7 September 1940, to Phyllis Agnes Robertson, aged 23. Johannesburg, 30 October 1951, to Marjorie Eileen Dunlop, aged 34. No death, no burial, no estate, no baptism, under any spelling.

What would fall outside it: AND HERE IS WHY IT PROVES LESS THAN IT LOOKS, WHICH IS THE POINT OF WRITING IT DOWN. FamilySearch's South African civil death registration is a NAME INDEX TO 1966: exact-year counts give 61,899 for 1960, 74,131 for 1965, 59,263 for 1966 - and then 178 for 1967 and nothing after it. Leonard died in 1973 or 1979. HIS ABSENCE FROM AN INDEX THAT ENDS SEVEN YEARS BEFORE HE DIED IS NOT EVIDENCE ABOUT HIM. The same bound removes Kurt, who died in 1968, and Henry Robert, who died in September 1969 and is nonetheless in the collection - as a BURIAL, in the Johannesburg cemetery registers, which run to 2019. So the negative that does carry weight is the narrower one: he was not buried in a Johannesburg cemetery, in a set that holds his father, his mother and two later Wears.

There is no post-1970 South African marriage register to reach by name on FamilySearch

Searched: "South Africa, Civil Marriage Records, 1801-1974", collection 2821281 - 2,556,837 records and 2,692,991 images. Counted by exact year rather than trusting the title.

Result: 1950 = 133,215. 1955 = 148,400. 1959 = 171,201. 1960 = 178,037. 1961 = 124. 1962 = 37. 1963 = 66. 1965 = 190. 1966 = 6. 1967 = 0. 1968 = 2. 1969 and after = 0.

What would fall outside it: A CLIFF, NOT A DECLINE. The collection is titled 1801-1974 and is, as a name index, 1801-1960. The images may well run to 1974; nothing lets you find one by name. THE GENERAL LESSON, AND THIS ARCHIVE HAS NOW LEARNED IT TWICE IN A WEEK: a collection's title describes the records, not the index, and an index that stops looks exactly like a family that leaves no record. The first time was the electoral registers, where a search box returned 3,478 rows and not one name.

Leonard did not die in England or Wales between 1970 and 1985 16 Sep 2026

Searched: FreeBMD (the GRO death index for England and Wales): surname WEAR, deaths, March 1970 to December 1985, NO forename given — 241 entries across 58 consecutive quarters *** WIDENED 22 SEPTEMBER 2026 WITH FREEBMD'S OWN PHONETIC FLAG. *** The original searched WEAR as an exact string. Soundex WEAR, deaths, Mar 1970 to Dec 1985 returns 6,664 matches — TWENTY-SEVEN TIMES the exact-spelling pool, and the first search never saw any of it. Narrowed by the forename Leonard it comes to about forty entries across WEIR, WARE, WARR, WRAY, WHEAR, WYER, WERE, WROE, WERRY, WHURR, WIER and WERRA.

Result: Sixteen of the 241 were born between 1915 and 1919. NOT ONE is Leonard Aubrey Robert Wear; the only 1917 birth is a Sam Wear of Halifax. A separate search on the forename Leonard returns one man, Leonard Rigby Wear, born 10 October 1897, died Claro June 1970. *** AND THE WIDENED SEARCH DOES NOT DISTURB IT. *** Of the forty phonetic Leonards, EXACTLY ONE is a true WEAR — Leonard Rigby Wear, born 10 October 1897, died Claro, June 1970 — and NOT ONE OF THE FORTY WAS BORN IN 1917. Leonard Aubrey Robert Wear was born at Hitchin in the first quarter of 1917. FreeBMD prints a date of birth against every death registered after 1969, which is what makes this decisive rather than suggestive: it is not that no matching name was found, it is that every candidate name was found and every one of them is a different man with a different birth year.

What would fall outside it: AS CLEAN AS A NEGATIVE GETS, and the reasons are worth stating. Every one of the 58 quarters is marked >99% complete, and FreeBMD's own coverage chart shows 1970–1983 at 100%. From 1969 the GRO death index carries FULL FORENAMES AND THE DATE OF BIRTH, so he could hide behind neither an initial nor a common surname. What it does NOT cover: deaths abroad, which are in the consular and overseas registers and have not been searched; and Scotland and Northern Ireland, which have separate registers. *** WHAT THE WIDENING DOES NOT FIX. *** FreeBMD is not complete. Its own coverage pages say so, and a death it has not yet transcribed is invisible to any spelling. The negative is about the GRO index AS FREEBMD HAS IT, not about the GRO index.

The British consular and overseas registers cannot contain Leonard's death 16 Sep 2026

Searched: GRO consular death series coverage; gov.uk guidance on registering a death abroad; published lists of excluded countries

Result: Three separate exclusions, all pointing the same way. The consular death registers run 1849–1965, so both candidate years fall outside. The modern statutory service covers deaths “on or after 1 January 1983” and is optional. And South Africa appears on the published list of countries where a death cannot be registered with the British authorities at all.

What would fall outside it: The first two are FIRM — a series end-date and gov.uk's own words. The third is PROBABLE, not proven: it is a local authority's summary, not legislation, and is marked as such. Any one of the three would be enough on its own.

Leonard has no deceased estate in South Africa 15 Sep 2026

Searched: National Archives (NAAIRS), all repositories, surname WEAR, 1970–1976

Result: No Wear deceased estate in the Transvaal at all between 1970 and 1976. The last is his father's, MHG 10245/69.

What would fall outside it: THE ESTATE INDEX STOPS AT THE END OF 1976 — 2,246 estate records in 1976, one in 1977, zero after. So 1973 is inside the index and he is not in it; 1979 is three years past its last year and cannot be tested this way at all. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page.

Leonard is not buried in any Johannesburg cemetery the index covers 15 Sep 2026

Searched: FamilySearch personas, f.collectionId=4453927 (Johannesburg cemetery registers, 1840–2019), surname WEAR, all 517 hits

Result: Six distinct Wears: Henrietta 1909, Henry Robert 1969, Martha Maud 1986, Edwina Ann 2014, Michelle Ann 2017, Dan B undated. NO LEONARD, under any spelling of his forenames. He bought grave 56045 and was not buried in it.

What would fall outside it: This is the INDEXED layer of ONE collection. It does not cover unindexed images within it, other Johannesburg cemeteries, or the Chevrah Kadisha — a separate database, and the one holding Kurt, Pauline and Salkow. A Wear would have no business in a Jewish cemetery, but the exclusion is stated rather than assumed.

Leonard Aubrey Robert Wear was granted no probate in England and Wales between 1970 and 1981 — and it took three wrong answers before this archive actually read the calendar 15 Sep 2026

Searched: probatesearch.service.gov.uk (Find a Will). ORIGINALLY: surname WEAR, year of death 1973, keyword AUBREY. RE-EXAMINED 22 September 2026, three times, because the first two re-examinations were both wrong. *** WHAT FINALLY SETTLED IT WAS READING THE REQUEST INSTEAD OF THE PAGE. *** The service answers in two halves. The DIGITAL half holds only grants issued from 1996 and says so, so a 1973 query returns 0 before anything is searched. The IMAGE half calls an OCR full-text endpoint, pp_cowindex_default_search, over the scanned calendar broken into chunks named for the first surname on them. CALLED DIRECTLY, THAT ENDPOINT PLAINLY WORKS: WEAR 1973 returns 41 chunks, SMITH 1973 returns 469, and LUWINSKI and the nonsense string ZQXVWYRK return 0. Two of the 41 bracket the name alphabetically — 009223_weake_1973.pdf (weake to weare) and 009225_wearmouth_1973.pdf (wearmouth to weatherby). *** AND THEN, THE SAME EVENING, THE CALENDAR WAS ACTUALLY READ. *** The chunk's OCR text is served as a property, textract:docContent, so the page never has to be rendered. Four chunks were read in full: 009223_weake_1973.pdf (Weake to Weare, so the whole of WEAR), 009474_wear_1974.pdf, 008820_weale_1979.pdf and 009151_wear_1980.pdf. THE TWO FOLLOWING YEARS WERE READ BECAUSE THE CALENDAR IS ORDERED BY YEAR OF GRANT, NOT YEAR OF DEATH — the 1973 volume contains people who died in December 1972. Separately the forename pair “LEONARD AUBREY” was put to every year from 1970 to 1981. CONTROLS FIRST THIS TIME: “Ernest Leslie” in 1973 returns 18, the nonsense string returns 0, and WEAR/SMITH/LUWINSKI return 41/469/0 — the count moves with the name.

Result: SIXTEEN WEAR GRANTS ACROSS THE FOUR YEARS, READ NAME BY NAME, AND NOT ONE IS HIM. 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. *** AND THE WIDER TEST IS THE STRONGER ONE: *** “Leonard Aubrey” across 1970-1981 returns TWENTY-SIX MEN — 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. NOT ONE IS A WEAR.

What would fall outside it: WHAT IT DOES NOT SETTLE: an English grant needs English assets, so a man who died in South Africa need never appear in this calendar at all. THIS IS A NEGATIVE ABOUT A ROUTE, NOT ABOUT A LIFE. And it rests on OCR — the very page that was read prints “Milliam” for William and “Minifred” for Winifred — WHICH IS WHY THE SURNAME BLOCKS WERE COUNTED OFF THE PAGE RATHER THAN RETURNED BY A QUERY. *** AND THE ROAD HERE IS KEPT BECAUSE IT IS THE USEFUL PART. *** This archive published three wrong answers about this one negative in a single evening: a zero read from the half of the service that structurally cannot hold a 1973 grant; then a control that had not been repeated; then a printed count — “41 document pages returned” — that turns out to be a defect in the service's own page, identical for every surname, for nonsense, and for both years. THE FAULT BEHIND ALL THREE WAS READING THE RENDERED PAGE INSTEAD OF THE REQUEST UNDERNEATH IT. What finally worked was calling the endpoint, running a positive and a null control before believing anything, and then reading the actual words on the actual page.

Ivone never naturalised as a South African 15 Sep 2026

Searched: Government Gazette full text, year by year 1946–1980, forename IVONE; plus the exact forms “Salkow, Ivone”, “Ivone Salkow”, “LUWINSKI, Ivone”, “MOTA FIGUEIREDO”, “Rocha, Ivone”

Result: Ten South African gazettes carry the forename in that window. All ten were opened and read; none is her. The exact forms return zero each.

What would fall outside it: *** CONTROLLED 22 SEPTEMBER 2026, AND IT HAD NOT BEEN. *** This negative stated its coverage limits but never showed that the corpus carries naturalisation notices at all. It does, in volume: THE WORD NATURALIZATION APPEARS IN 28,470 SOUTH AFRICAN GAZETTE DOCUMENTS on gazettes.africa. The archive's own work on Kurt's naturalisation also read 246 gazettes carrying the heading, in both series — intention to apply, and lists of certificates issued — in full for 1975 and 1976. So a search of 1946-1980 under her forename and three surnames is a real search of a populated class, not a query into a gap. Covers the notice of intention and the issued-certificate lists, under all three of her surnames. It does not cover a naturalisation granted outside the gazette's digitised run, and the OCR of two-column lists is poor enough that one entry this evening had to be unpicked by hand.

Kurt never naturalised as a South African 15 Sep 2026

Searched: Government Gazette full text, every year, surname LUWINSKI

Result: Exactly eight hits across the whole archive and this archive had already read all eight — six Piccadilly Hat Company partnership notices, the 1972 Paula application, the 1975 Pallou's Dress Centre notice. None is a naturalisation.

What would fall outside it: STRONGER THAN THE WEAR NULLS, because Luwinski is a rare surname and the same search demonstrably finds him eight times. It does not rule out a naturalisation recorded only in a file and never gazetted. *** 11 OCTOBER 2026 — RE-RUN ON THE NAAIRS SUCCESSOR UNDER ELEVEN SPELLINGS AND IT HOLDS. *** LUWINSKI 7 · LUWINSKY 1 · LEWINSKI 5 · LOWINSKY 7, and LUWINSKA, LOWINSKI, LOWINSKA, LUWINSKIE, LUVINSKI, LOWINSKE and LUWINS all zero — with SMITH and SCHMIDT as live controls. NO NATURALISATION FILE AND NO IMMIGRATION FILE UNDER ANY OF THEM. LUWINSKY, the spelling this archive knew the South African record used and had never searched, returns EXACTLY ONE RECORD and it is the 1948 divorce file already held. NARSSA's own page names the portal used as the COMPLETE database.

There is no immigration file for Kurt 15 Sep 2026

Searched: NAAIRS, all repositories, IMMIGRATION+LUWINSKI and PIO+LUWINSKI

Result: Zero. The only Wear immigration file is KAB PIO 907 ref 38964E, “Immigration papers H C Wear”, 1949 — wrong initials, wrong decade.

What would fall outside it: NAAIRS indexes the PIO series at item level, so this is a real search of a real series. It does not cover series that were never indexed, nor records still with Home Affairs. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page. *** 11 OCTOBER 2026 — RE-RUN ON THE NAAIRS SUCCESSOR UNDER ELEVEN SPELLINGS AND IT HOLDS. *** LUWINSKI 7 · LUWINSKY 1 · LEWINSKI 5 · LOWINSKY 7, and LUWINSKA, LOWINSKI, LOWINSKA, LUWINSKIE, LUVINSKI, LOWINSKE and LUWINS all zero — with SMITH and SCHMIDT as live controls. NO NATURALISATION FILE AND NO IMMIGRATION FILE UNDER ANY OF THEM. LUWINSKY, the spelling this archive knew the South African record used and had never searched, returns EXACTLY ONE RECORD and it is the 1948 divorce file already held. NARSSA's own page names the portal used as the COMPLETE database.

NAAIRS cannot answer whether anyone in this archive naturalised 15 Sep 2026

Searched: NAAIRS central-government database, keyword NATURALIZATION

Result: 342 documents, and every one is a policy file or a bill — the 1926 Act, its 1927 amendment, South West Africa bills.

What would fall outside it: INDIVIDUAL NATURALISATION FILES ARE NOT INDEXED AT ITEM LEVEL. So the nulls for NATURALIZATION+WEAR and NATURALIZATION+LUWINSKI are worth nothing at all, and are recorded here only to stop anyone running them again. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page.

The 1986 reopening of grave 56045 was not recorded 15 Sep 2026

Searched: West Park burial register folio W 104 and the Private Graves Register index folio W 8, rebuilt from their deep-zoom tiles

Result: Martha Maud's Remarks column is empty, not illegible. The private-graves index holds exactly one Wear, her husband.

What would fall outside it: A second interment into an existing private grave generated no grave transaction, so no authority was ever written down. The record was not lost; IT WAS NEVER MADE. What is untested is the undertaker's own file.

The South African civil marriage register never asked for parents 15 Sep 2026

Searched: The 1940, 1948, 1951, 1952, 1953 and 1956 registers, read at full resolution

Result: Two form designs across the period, and neither carries a parents' column that was ever filled in. The 1956 form has one, ruled through and empty.

What would fall outside it: BOUNDED 1940 TO 1970. Nobody has checked whether the form changed afterwards, and there are three post-1970 Wear marriages to test it on.

Krmpote has no birth register before 1879 15 Sep 2026

Searched: FamilySearch catalogue, and the Croatian State Archives register search (HR-HDA-1448)

Result: Both give Krmpote births starting 1879. Two independent indexes agree.

What would fall outside it: Neither covers books that were never filmed. The parish's own status animarum, XIX–XX, is digitised and has NOT been read — it is a household census with birth dates and would not be caught by a birth-register search.

Leonard never appears in the Government Gazette — A WEAK NULL, recorded as weak 15 Sep 2026

Searched: Full text, every exact form of his name; and loose sweeps on Wear with engineer, Fairlands, Johannesburg

Result: Exact forms return zero. Loose forms return 3,469 and 5,568 documents, none of them him.

What would fall outside it: WEAR IS AN ORDINARY ENGLISH WORD — it appears in trade names, wool-labelling prosecutions and railway reports about flange wear. A null on this surname is nearly worthless, and is the opposite of the LUWINSKI null, which is strong precisely because the surname is rare and the same search finds him eight times. The two are not the same kind of object.

None of Leonard's three dead sisters has a gazetted estate 16 Sep 2026

Searched: Government Gazette full text: “CYNTHIA YVONNE ELIZABETH”, “JANETTE EILEEN MAUD”, and “KATHLEEN EDITH” with FOOT across 2009–2014

Result: Cynthia returns nothing at all. Janette returns only her own 1975 naturalisation notice and the 1976 certificate list. Kathleen returns nothing in her death window.

What would fall outside it: *** CONTROLLED 22 SEPTEMBER 2026, AND THE CONTROL NEARLY WENT THE OTHER WAY. *** The phrase “Estates of deceased persons” returns 1,219 South African documents across the run but NOTHING AT ALL for 2009 to 2013 — which looked at first like a coverage hole that would have voided this negative. IT IS NOT: the heading simply changed wording. “MASTER OF THE HIGH COURT” returns 6,594 documents, with 59 in 2009, 74 in 2010, 67 in 2011, 133 in 2012, 172 in 2013 and 164 in 2014. THE YEARS ARE IN THE CORPUS AND THE ESTATE MACHINERY IS IN THEM. So the three women were searched for in gazettes that do carry estate business, and they are not there. THIS IS EXPECTED, NOT SUSPICIOUS. A Form J193 estate advertisement is required only above a value threshold, and three women who died in Johannesburg suburbs in 2000, 2001 and 2010 would very probably fall under it. It means the estates cannot be ordered by number and must be asked for by name and date of death — not that they do not exist.

There is no Maud Mabel Castle in the records — A WEAK NULL 16 Sep 2026

Searched: FamilySearch, full forename pair MAUD MABEL with surname CASTLE, England and South Africa; and NAAIRS for LONGHURST

Result: No record anywhere for a Maud Mabel Castle. The 136 NAAIRS Longhursts are overwhelmingly Cape-based and older, with nothing joining any of them to a Castle.

What would fall outside it: FamilySearch fuzzy-matches forenames badly — it collapses Janette into Jane, Jean and even Ware — and its England coverage is not complete. This does not dispose of the 1915 and 1919 Johannesburg divorces, which name her plainly. The real tests are two orderable documents: MHG 3632/50 and WLD 155/1915. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page.

No Leibholz appears anywhere in the South African archives 16 Sep 2026

Searched: NAAIRS, all repositories and national registers, surname LEIBHOLZ — and, on 11 OCTOBER 2026, RE-RUN UNDER SIX SPELLINGS on the service's successor: LEIBHOLZ, LEIPHOLZ, LEIBHOLTZ, LEIPHOLTZ, LAIBHOLZ, LEIBHOLS, with SMITH (47,936) and SCHMIDT (3,418) as live controls.

Result: Zero documents. *** AND STILL ZERO UNDER ALL SIX SPELLINGS, 11 OCTOBER 2026. *** LEIBHOLD returns 1 and is a different surname.

What would fall outside it: Expected rather than informative — the Leibholz brothers were in Berlin and East Prussia and none of them reached South Africa. Recorded so that nobody runs the search again on the assumption it was never tried. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page. *** 11 OCTOBER 2026 — THE SERVICE IS BACK AND THIS BOUND IS SETTLED RATHER THAN ARGUED. *** NAAIRS returned 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. AND NARSSA'S OWN PAGE STILL CARRIES THE TWO-DATABASE WARNING AND NAMES THAT PORTAL AS THE COMPLETE ONE — “To search the complete database please click on: Search NAAIRS Database →”, whose href is exactly that URL. So for the 11 October re-run the question “which twin answered” is answered from the archive's own words, not from the vocabulary of a results page. THE NEGATIVE IS NOW WIDE (six spellings), CONTROLLED (two live controls) AND ATTRIBUTED (the database NARSSA calls complete), and it should no longer be listed as provisional. The 15 and 16 September searches remain unattributable, which is unchanged and is a record-keeping fault, not a doubt about the answer.

No Johannesburg street directory of the 1940s or 1950s is reachable without a reading room 16 Sep 2026

Searched: ancestors.co.za street directories; eGGSA; FamilySearch South Africa online records guide

Result: The only South African street directory digitised and searchable is PIETERMARITZBURG 1897. Voters rolls online cover 1925 and 1931 only, across 148 towns.

What would fall outside it: This is about ONLINE reach, not existence — the Johannesburg directories exist on paper and in libraries. And the 1925/1931 voters rolls, useless for the Wears who arrived in the late 1930s, are exactly right for the Momple, Robertson and Longhurst families, who were Johannesburg people in those years.

No spelling of Ivone's maiden name appears in a South African naturalisation notice, 1930–1960 16 Sep 2026

Searched: gazettes.africa full-text, South African issues only, every year from 1930 to 1960 inclusive. FIRST PASS searched three spellings — FIGUEIREDO, DE FIGUEIREDO, MOTA FIGUEIREDO. SECOND PASS, run the same day after David pointed out that she is recorded under several names, added the other three the records actually use — FIGUEREDO (her own 1953 marriage register), FIGUERIDO (WLD 2225/1953) and FIGARETTO (the NAAIRS index itself) — plus FIGUEIRIDO, and the married surnames LUWINSKI, LUWINSKY and LUWENSKI.

Result: NO NATURALISATION NOTICE UNDER ANY SPELLING. The first pass returned five South African hits in thirty-one years, two of them deceased-estate notices new to this archive: ESTATE 7029/54 — FIGUEIREDO, AGOSTINHO VICENTE SEBASTIAO, of Johannesburg, account advertised 22 July 1955; and ESTATE 2910/60 — MONTEIRO, JOSÉ DE FIGUEIREDO, OF LOURENÇO MARQUES, PORTUGUESE EAST AFRICA, advertised 2 December 1960. THE SECOND PASS ADDED NOTHING. Figueredo returns five South African hits across 1940–1943 and ALL FIVE ARE THE SAME PERUVIAN TRADING ADDRESS in a wartime enemy-trading blacklist — "Figueredo 792, Callao, Peru". Figuerido returns two hits on the whole continent and none in South Africa. FIGARETTO, FIGUEIRIDO AND LUWENSKI RETURN NOTHING ANYWHERE IN ANY AFRICAN GAZETTE — they are index corruptions, not spellings anyone ever printed. Luwinski and Luwinsky return the same eight documents already known.

What would fall outside it: TWO CONTROLS, WHICH IS WHAT MAKES THIS WORTH ANYTHING. (1) THE CORPUS CARRIES NATURALISATION MATERIAL: searched for "certificate of naturalization", "naturalization" and "Aliens Act" in 1938, 1946, 1950 and 1955 — South African hits in every year tested, up to ten a year on a capped result set. (2) THE SEARCH IS NOT DEFEATED BY SPELLING: six recorded forms of the maiden surname and three of the married one were each tried across all thirty-one years. THE FIRST VERSION OF THIS NEGATIVE OVERSTATED ITSELF by searching only three spellings while this archive's own notes record six — that is corrected here rather than left looking stronger than it was. WHAT WOULD STILL FALL OUTSIDE IT: (a) THE OCR ON THESE SCANS IS DREADFUL — one 1943 hit renders as "Epo .Figueiredo,CarlosDe" — so this negative is SOFT in a way the FreeBMD one is not, and no list of spellings fixes a scanner that loses the word altogether; (b) it searches the published notice, not the file, and not every naturalisation was gazetted under every statute; (c) IT DOES NOT TOUCH THE SAB NATURALISATION AND ALIEN-REGISTRATION FILES THEMSELVES, which are indexed on NAAIRS and were not searched because NAAIRS was unreachable on the day; (d) if she took her husband's nationality by marriage in February 1948, as the law before September 1949 allowed, THERE WAS NEVER A NOTICE TO FIND.

WITHDRAWN — ‘NAAIRS is down’ was wrong. The URL had moved and this archive was holding a stale one 16 Sep 2026

Searched: http://national.archsrch.gov.za/sm300dl, the entry point this archive had used throughout and had never re-checked against the National Archives portal that publishes it.

Result: THE HOST DOES NOT RESOLVE AND NEVER WILL AGAIN — that part was accurately observed. THE CONCLUSION DRAWN FROM IT WAS WRONG. The portal at national.archives.gov.za/naairs.htm still publishes a working Search link, and it points somewhere else entirely: http://www.national.archsrch.gov.za/sm300cv/smws/sm300dl — a WWW PREFIX AND A PATH PREFIX this archive did not have. That URL returns HTTP 200, issues a session token and searches normally. FIGUEIREDO returns 43 documents, which is exactly the figure this archive recorded on 14 September.

What would fall outside it: RECORDED AS A FAILURE RATHER THAN QUIETLY DELETED, because the failure is the instructive part. THE OBSERVATION WAS SOUND AND THE INFERENCE WAS NOT: a dead hostname was read as a dead service, when the service had simply moved and this archive was the only party still using the old address. It took DAVID OPENING THE PUBLISHED PORTAL PAGE IN A BROWSER to find it — thirty seconds of looking at where the link actually points, against an afternoon of diagnosing DNS. THE COST WAS NOT ONLY TIME. The draft reply to Ms Xaba had a paragraph telling her, helpfully, that her own institution's search system was down. IT WOULD HAVE BEEN WRONG, and it was removed before the letter was sent. THE RULE TAKEN FROM IT: when an endpoint this archive depends on stops answering, RE-READ THE PAGE THAT PUBLISHES THE LINK before concluding anything about the service behind it.

The variant spellings each return exactly one NAAIRS record, and the correct spelling returns none of them 16 Sep 2026

Searched: NAAIRS, All Archives Repositories (RSAE), on the corrected endpoint, 16 September 2026. Searched FIGUEIREDO, FIGARETTO, LUWENSKI, LUWINSKY as separate queries, and the full record fetched for each.

Result: FIGUEIREDO — 43 documents, the same 43 recorded on 14 September. FIGARETTO — 1: ILLIQUID CASE RESTITUTION OF CONJUGAL RIGHTS IVONE LUWENSKI (FORMERLY DE WITT, BORN FIGARETTO) VERSUS KURT, which is TAB WLD 2545/1950. LUWENSKI — 1: the same file. LUWINSKY — 1: TAB WLD 1950/1948. EACH VARIANT SURFACES A FILE THAT THE CORRECT SPELLING DOES NOT RETURN. WLD 2545/1950 is the only action in which Ivone is the plaintiff, and it is reachable ONLY through two spellings nobody would choose to type.

What would fall outside it: THE FIRST TWO ATTEMPTS AT THIS SEARCH RETURNED ZERO AND WERE NOT BELIEVED, which is the only reason the result is worth anything. The first posted an incomplete form and NAAIRS answered, in words, 'the search was bypassed as no search word was entered' — a message that would have read as an empty holding to anyone parsing for row counts. The second returned zero on a run that had returned one a minute earlier, and was treated as throttling rather than data. THIS ARCHIVE HAS TWICE BEFORE RECORDED FALSE EMPTINESS FROM EXACTLY THIS FAILURE and the rule now is that a zero from a flaky interface is not a finding until it reproduces. WHAT WOULD STILL FALL OUTSIDE IT: only four spellings were put to NAAIRS, not all fourteen recorded on the names page; the search was RSAE, the all-repositories database, which should but may not subsume every depot; and NAAIRS indexes file titles, not file contents, so a woman named inside a document she is not a party to remains invisible to any spelling. || *** AND A BOUND ADDED 27 SEPTEMBER 2026 THAT IS NOT ABOUT THE INDEX BUT ABOUT THE DATABASE. *** Every bound this archive wrote for its NAAIRS nils concerns THE INDEX'S COVERAGE OF THE RECORDS — which series are indexed at item level, where the estate index stops. NONE CONCERNS THE DATABASE'S COVERAGE OF THE INDEX, because on 15-16 September there was no reason to think there were two databases. THERE ARE. NARSSA'S OWN SEARCH PAGE STATES IT, verified at source 27 September: "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." *** WHAT THIS ARCHIVE CAN SHOW ABOUT WHICH ONE ANSWERED, AND IT IS EVIDENCE NOT PROOF: *** the vocabulary of these searches is the OLD database's — "all repositories", "national registers", "central-government database", the repository code RSAE — and the new fulltext search has no such repository selection. The 16 September searches name the endpoint www.national.archsrch.gov.za/sm300cv/smws/sm300dl, which is the classic Search Manager service, i.e. THE OLD COMPLETE ONE. Probed again 27 September: it answers 503, and the old portal page naairs.htm is now 404. SO THE BALANCE OF EVIDENCE IS THAT THE OLD COMPLETE DATABASE ANSWERED. IT IS NOT ESTABLISHED FOR THE 15 SEPTEMBER SEARCHES, whose pre-move entry point was never written down as a URL. THE NIL STANDS AND IS NOW BOUNDED: it was taken from a service that had a partial twin, and this archive did not check which twin it was talking to. Raised by a peer session, verified here at NARSSA's own page.

WITHDRAWN WITHIN THE HOUR — the archives index does not list naturalisation files AT ALL, so their absence proves nothing 16 Sep 2026

Searched: First pass: the National Archives index on all eleven tested spellings of Ivone's maiden surname, and separately on LUWINSKI, LUWINSKY and LUWENSKI for Kurt — every depot, every series, no date limit. THEN, ON DAVID'S QUESTION ABOUT KURT, A CONTROL TEST THAT 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. So the index was searched for him.

Result: NO NATURALISATION FILE FOR OPPENHEIMER EITHER. A man this archive can PROVE naturalised, with the gazette notice in hand, returns three records and not one of them is his naturalisation. THE INSTRUMENT CANNOT DETECT A KNOWN POSITIVE. Therefore the absence of a Figueiredo or a Luwinski naturalisation file in that index says NOTHING WHATEVER about whether such files exist. The first version of this entry concluded that no naturalisation record exists. IT WAS WRONG, and it was wrong in the most ordinary way: it mistook the limits of a finding aid for a fact about the world.

What would fall outside it: WHAT SURVIVES, AND IT IS STILL WORTH HAVING. The surname searches DID return sixteen immigration files at KAB — twelve IRC, four PIO — and none at Pretoria, which is real evidence about WHERE the Portuguese community is indexed even though it is no evidence about naturalisation. WHAT THIS NOW MEANS IN PRACTICE: naturalisation files are evidently not item-indexed online, so THE ONLY ROUTE IS SOMEBODY STANDING IN THE READING ROOM ASKING WHETHER A NATURALISATION REGISTER OR CARD INDEX EXISTS. That question is now in both outstanding letters — to the Reading Room supervisor, who would know what finding aids exist, and to the researcher, framed as a question rather than a search. AND THE LAW STILL MAY HAVE PRE-EMPTED IT FOR IVONE: she 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.

The two Giddins DNA matches do NOT connect through the Hertfordshire Giddinses

Searched: On 16 September 2026 the Castle line was traced to JAMES GIDDINS of Norton End, Baldock, from the 1861 and 1871 censuses. The same afternoon, the archive's own guard flagged GIDDINS as a surname carried by two living DNA match holders — 43.36 cM and 62.89 cM, both in cluster 12. The larger holder's public tree, eighty-nine people, was then opened and read.

Result: IT IS A DIFFERENT FAMILY. That tree's GEORGE GIDDINS, 1846–1890 — whose dates sit within a year of James's son George, born about 1847 — was born on 17 JULY 1846 AT WITHYBROOK, WARWICKSHIRE, and died in the Wandsworth district of London. HIS FATHER IS RICHARD GIDDINS, NOT JAMES. The other surnames in that tree are Ratcliffe, Conway, Woodhams, Wall, McMellon and Prescott — none of them Hertfordshire. THE SURNAME IS NOT THE CONNECTION.

What would fall outside it: WHAT THIS DOES AND DOES NOT SAY. It does NOT say the matches are false: 62.89 cM is 62.89 cM and both holders are genuinely related to Derrick. It says only that THE REASON IS NOT THE GIDDINS SURNAME — which should never have been assumed, because a match holder's surname is one of eight great-grandparental lines and there is no reason it should be the relevant one. RECORDED BECAUSE OF HOW NEARLY IT WENT THE OTHER WAY. Three paragraphs had already been written calling these 'the first DNA to touch the Castle line above George William Castle', with the fourth-cousin arithmetic worked out and fitting. THE ARITHMETIC WAS SOUND AND THE PREMISE WAS NEVER CHECKED. Opening the tree took four minutes. WHAT REMAINS OPEN: the smaller holder's tree runs to twelve people and is too thin to read; and the Hertfordshire Giddinses of Norton End are still untouched by any DNA at all.

No Luwinski, in any spelling, appears on any South African page FamilySearch has read — but the pages it has read stop in 1931

Searched: FamilySearch full-text search, which reads the words on unindexed images rather than a name index, signed in, 17 September 2026: LUWINSKI, LUWINSKY, LUWENSKI, and the phrase 'Piccadilly Hat' — the name of Kurt's own shop at 81 Eloff Street

Result: Thirty-three hits across the three spellings and EVERY ONE IS AMERICAN — Milwaukee and Chicago naturalisation petitions, New York and Philadelphia passenger lists, Ohio, Minnesota and Massachusetts vital records. The shop name returns nothing but Lancashire trade directories of the 1890s.

What would fall outside it:

There is no estate notice for Leonard Aubrey Robert Wear in the South African Government Gazette — and no Phyllis Agnes Norquoy either

Searched: gazettes.africa, free and full-text across 54,313 South African issues, 17 September 2026. The two searches that matter are WEAR AND AUBREY, which returns 38 documents, and NORQUOY, which returns 41. EVERY ONE OF THE SEVENTY-NINE WAS READ. Also run: Wear and Norquoy, Wear and Kwele, Norquoy and Phyllis, Norquoy and Agnes, Luwinski and Ivone, Luwinski and Frances, Hostowsky, Salkow and Ivonne.

Result: No Leonard. No Phyllis. An estate notice under the Administration of Estates Act is the one document that would carry his date of death and settle whether it was 1973 or 1979, and it is not there under a search that would catch it if both words survived the scanner.

What would fall outside it:

WITHDRAWN — the England & Wales Electoral Registers 1832-1932 cannot be searched for a person at all, and this archive said it could

Searched: Findmypast, that record set, 18 September 2026: WEAR with a Bermondsey place filter; WILLIAM WEAR with no filter; JOHN SMITH as a high-frequency control; LUWINSKI as a rare-surname control; and a nonsense string as a null control.

Result: EVERY ROW THE SET RETURNS SAYS 'NO NAME ON RECORD'. William Wear returns 3,478 of them and the locations are DEVON and DURHAM — Weare in Devon, the River Wear in Durham. John Smith returns the same nameless rows in Lancashire. Luwinski returns zero, and so does a nonsense string. The set is page images of volumes; it indexes no elector, and what the name box matches is a place.

What would fall outside it:

The record that would name Georg Leibholz's parents does not exist — Hammerstein's civil registers and its synagogue books are both gone

Searched: Archiwum Panstwowe w Koszalinie, asked on 14 September 2026 for the registers of the former German civil registry office at CZARNE — Standesamt HAMMERSTEIN, Kreis Schlochau — for 1847 and 1884, and for the books of the Jewish community there. Answered 16 September 2026, reference O-S.6342.406.2026, over the signature of Krzysztof Skrzypiec, Head of the Slupsk Branch.

Result: NEITHER SURVIVES IN THAT ARCHIVE. No Standesamt Hammerstein registers for 1874 TO 1899 — the whole span of German civil registration up to the end of the century — and NO SYNAGOGUE-COMMUNITY BOOKS FOR CZARNE AT ALL. The letter goes further than a simple nil return and says they hold no information about the fate of either series, or where it might now be kept.

What would fall outside it:

Frederick James Castle has no surviving First World War service file — and that is what a career soldier looks like, not a gap

Searched: Findmypast, BRITISH ARMY SERVICE RECORDS — which carries WO 363, the 'Burnt Documents', and WO 364, the pension claims — surname CASTLE, forename FRED*, all 78 results read across four pages, 17 September 2026.

Result: NOT ONE ENTRY FOR THE SCOTTISH RIFLES, THE CAMERONIANS, OR SERVICE NUMBER 10061. There are Fred James Castles in the set — 98411 Labour Corps, 10524 Devonshire Regiment, 40524 13th Reserve Cavalry — and every one of them belongs to another man; the Labour Corps one is the Royal Berkshire soldier whose medal card gives 33601 and 98411 together.

What would fall outside it:

Frederick James Castle did not die on army service abroad and did not marry in an army chapel abroad — under that name

Searched: Findmypast, 17 September 2026. BRITISH ARMED FORCES AND OVERSEAS DEATHS AND BURIALS: surname CASTLE, forename FRED*, all ten results accounted for. BRITISH ARMED FORCES AND OVERSEAS BANNS AND MARRIAGES: surname CASTLE, ALL FIFTY-NINE RESULTS READ ACROSS THREE PAGES. These are the registers a soldier of the 1920s appears in instead of the ordinary General Register Office series.

Result: NOTHING IN EITHER. No Scottish Rifles death and no number 10061; the nearest by name is FREDERICK J CASTLE who died in 1917 and is PRIVATE 47783 OF THE NORTHUMBERLAND FUSILIERS, buried at Roclincourt Valley Cemetery in France. In the marriage set there is NOT ONE Fred, Frederick or Fredk among fifty-nine Castles, who marry instead at Malta, Kolar, Istanbul, Singapore, Italy, Aldershot and Johannesburg.

What would fall outside it:

The name LUWINSKI appears in no record described by any archive in The National Archives' Discovery catalogue

Searched: discovery.nationalarchives.gov.uk, the catalogue of The National Archives AND of more than two thousand five hundred other British archives — county record offices, university and cathedral libraries, regimental and company collections. Searched 17 September 2026 for LUWINSKI and LUWINSKY, and RE-RUN 21 September 2026 across fourteen spellings: LUWINSKI, LUWINSKY, LUWENSKI, LUWINSK, LOWINSKI, LEWINSKI, and the eight Leibholz forms LEIBHOLZ, LEIPHOLZ, LEYBHOLZ, LAIBHOLZ, LEIBHOLTZ, LEIPHOLTZ, LEIBOLZ, LEIPHOLS.

Result: COUNT 0 for both. By contrast LEIBHOLZ returns sixteen, none of them ours — Gerhard Leibholz, Dietrich Bonhoeffer's brother-in-law, at Lambeth Palace Library; a 1948 naturalisation certificate; two Home Office alien files; and an internment card for a Tobias Leibholz born at Breslau in 1906. *** RE-RUN 21 SEPTEMBER UNDER THE EIGHTH SPELLING AND THE NEGATIVE HOLDS AND IS NOW WIDER. *** LUWENSKI and LUWINSK also return COUNT 0, so four spellings of the name are now bracketed rather than two. LEIPHOLZ returns 0, as do LEYBHOLZ, LAIBHOLZ, LEIBHOLTZ, LEIPHOLTZ, LEIBOLZ and LEIPHOLS — so the sixteen Leibholz records are the whole of what Discovery holds under any spelling of that name. TWO NEIGHBOURING SURNAMES DO RETURN, AND ARE NAMED HERE SO NOBODY HAS TO FIND THEM TWICE: LOWINSKI returns 20 (Thomas Esmond Lowinsky, the painter, 1892-1947) and LEWINSKI returns 228. Neither is this family and neither is claimed as one.

What would fall outside it:

The Bundesarchiv Gedenkbuch holds no Leibholz under any misspelling this archive had not already tried — and that turned out to be false

Searched: apps.bundesarchiv.de/gedenkbuch, the free name box, ten spellings put to it on 22 September 2026: LEIBHOLZ, LAIBHOLZ, LEIBHOLTZ, LEIPHOLZ, LEIBOLZ, LEBHOLZ, LEIHOLZ, LEIBHOLD, LEIBHOLCZ, LELBHOLZ. The control term LEIBHOLZ was re-run between the others four times and returned 50 every time, so every negative below sits between two proofs that the instrument was answering.

Result: SEVEN OF THE TEN ARE CLEAN ZEROS — Leibholtz, Leipholz, Leibolz, Lebholz, Leiholz, Leibholcz and Lelbholz. LEIBHOLD RETURNS ONE AND IT IS A FALSE POSITIVE: the entry is “Passau, Leibhold”, where Leibhold is the man's forename and Passau his surname. AND LAIBHOLZ RETURNS TWO PEOPLE THIS ARCHIVE HAD NEVER SEEN — Georg Laibholz born 1 January 1883 and Lidia Laibholz born 1884, both of Berlin, both dead at Auschwitz, and neither a duplicate of the Georg and the Lydia already among the fifty, who carry different birth dates and died at Kulmhof. One vowel had been hiding two murdered people from every search run here.

What would fall outside it: WHAT THE ZEROS ARE WORTH is set by the control, and the control is strong: the same box returned 50 for Leibholz immediately before and after the run. WHAT THEY ARE NOT is a statement about the Gedenkbuch's coverage — the memorial book is the murdered as the Bundesarchiv has been able to reconstruct them, and a name absent from it may simply never have been reconstructed. AND THE RUN CAUGHT A FAULT IN ITS OWN INSTRUMENT: PRESSING RETURN IN THAT BOX DOES NOT SUBMIT THE FORM. The first attempt entered five terms with the Return key and read back “50 Results” for every one — the previous page each time, because nothing had been sent. Five negatives that were never tested would have entered this archive as five findings. They were caught because 50 is not a believable answer for five different misspellings, and the run was done again with a click on the button, giving 0, 50, 0, 1 and 0. THE LESSON IS NOT ABOUT THIS FORM: a reading that does not CHANGE is not a reading, and a sweep that cannot tell a fresh answer from a stale page is measuring nothing.

Sally Leibholz's death cannot be found by searching the Kuyavia-Pomerania portal — the name is not indexed and the Szubin deaths have no alphabetical index

Searched: www.genealogiawarchiwach.pl, 22 September 2026. Three searches with their controls. (1) Town SZUBIN, category AKTA ZGONÓW — the death registers. (2) The same, category SKOROWIDZE ZGONÓW — the alphabetical death indexes. (3) Free text on the surname LEIBHOLZ against the surname SCHMUL as a control, SCHMUL being the bride's family from the same town and the same decades.

Result: (1) 190 DEATH UNITS across 20 teams — Szubin has THREE registry offices, Miasto and Wieś I and Wieś II, and all are digitised. (2) ZERO SKOROWIDZE ZGONÓW. Birth and marriage skorowidze exist for Szubin; death ones do not. (3) LEIBHOLZ RETURNS EXACTLY ONE SCAN in a 228,278-person index — the 1874 marriage — while SCHMUL RETURNS 119 SCANS FROM 34 TEAMS ACROSS 104 UNITS.

What would fall outside it: THE CONTROLS ARE WHAT MAKE THIS WORTH ANYTHING, AND THEY WERE RUN FIRST THIS TIME. The Skorowidze zero is not a broken filter: the identical facet with AKTA ZGONÓW returns 190. The single Leibholz is not a dead search box: the same box returns 119 for the bride's surname. So the portal's index is rich for Schmul and blind to Leibholz, and a Leibholz death at Szubin — if one exists — would have to be read off the page rather than found. *** WHAT THIS IS NOT IS EVIDENCE THAT HE DID NOT DIE AT SZUBIN. *** Nothing here has been read; 190 units stand unopened, and with no death year to aim at, a man born in 1847 could be in any of them. *** AND THE STRONGER POINT IS THAT SZUBIN MAY BE THE WRONG TOWN ALTOGETHER: *** this family moved again. Kurt was born in Berlin in 1912, the household is at Chodowieckistraße in 1913, and Jacob married in Berlin in 1924. If Sally went with his sons, he died in Berlin, and the instrument for that is the Landesarchiv Berlin name indexes this archive already reads fluently. Paging 190 Szubin units before testing Berlin is the expensive way round.

Friedrich Leipholz's death is not registered in Reinickendorf — his own borough — in any year from 1974 to 1994

Searched: Landesarchiv Berlin, P Rep. 130, Namensverzeichnisse zum Sterberegister, Standesamt Reinickendorf. The 1974-1981 volumes were read on 21 September 2026 and the thirteen remaining, 1982 to 1994 (Nr. 1478-1490), on 23 September, through pdf.js over HTTP range requests — 807 pages of 2,120, out of roughly 700 MB, with nothing downloaded. NOT A NAME SEARCH: the OCR mangles this surname in five of eleven volumes elsewhere, so the test is to print the whole alphabetical slot and read it, recording an absence only when a name above it and a name below it are both in hand. BOTH SPELLINGS WERE BRACKETED SEPARATELY — Leibholz falls between LEH- and LEIC-, Leipholz between LEIN- and LEIS-, and they are two different slots. The text layer was tested before the sweep rather than after: this series prints names ("Blank Franz Erich 11.07. 1652"), unlike the Kreuzberg volumes of the same period, and a frequency control on the first nineteen pages returned Becker 6, Berger 5, Brandt 8 and Schmidt 0 — right for pages that have not yet left the letter B.

Result: TWENTY-ONE CONSECUTIVE YEARS READ, ALL BRACKETED ON BOTH SIDES OF BOTH SPELLINGS, AND HE IS IN NONE OF THEM. He is in the Berlin telephone book at Schwabstraße 4 in Reinickendorf to 1973-74 and absent from the 1974-75 volume, so he leaves the record during 1974 and his own borough never registers his death. THE FIRST PASS OF THIS SWEEP WAS DISCARDED: it bracketed only the Leibholz slot and would have published thirteen negatives covering Leipholz too — in the 1990 volume the nearest name above Leibholz is LEIPE, which sorts below Leipholz, so that slot had never been shown to be read. Two volumes, 1992 and 1993, failed mid-read with network errors and were not counted as absences; both read clean on retry. THE SWEEP'S ONE POSITIVE: Marie Leipholz, Sterberegister 1988 Nr. 2306, the only Leipholz in all twenty-one years, dying in the borough he lived in.

What would fall outside it: A death outside Reinickendorf. He may have moved: the Landesarchiv's full catalogue puts the whole of "some other Berlin borough" at 296 death-index volumes across 25 offices — bounded, and no longer a vague elsewhere. Or he may have died after 1994, which would make him about ninety-nine and puts him past the thirty-year closure. Or he may simply have given up the telephone and lived on, which at eighty is not unusual: HEINRICH LEIBHOLZ VANISHED FROM THE BERLIN DIRECTORY AFTER 1942 AND DID NOT DIE UNTIL 1948. Leaving a telephone book is not dying, and nothing here says he died in 1974 — only that he stopped being listed.

The rule this page is trying to enforce

Do not publish a null you did not bound. Twice this evening a search returned nothing and the nothing was the archive’s own fault: once because a page’s results were still loading, and once because a form field was posted empty. Both looked exactly like an empty holding. Neither was.

So a negative enters this archive with three things attached or it does not enter at all: what was searched, over what range, and what would still fall outside it.