Governance
Corrections and changelog
A record with no corrections in it is either very new or is not being read by anyone who knows better. This one is very new.
5 logged changes2 corrections0 superseded claims retained0 accepted contributions
Nothing on this record is silently corrected. When a claim turns out to be wrong it is superseded, not deleted: the old claim keeps its identifier, is marked as superseded, and stays reachable — so anyone who cited it can see what happened to it, and nobody repeats the reasoning that produced it. The same applies to a negative finding, which is recorded precisely so the next researcher does not spend the same afternoon on the same dead end.
A change to the canonical record cannot be made without a logged reason. That is a structural constraint rather than a habit — the reason field is required by the schema and the build refuses a change without one. Corrections are accepted from anyone, are reviewed before they enter the record, and are credited to the person who found the problem, by name, permanently. No externally credited correction has been logged yet; the first one will be named here.
No external contributions have been received yet. The most useful thing anyone can send this project is evidence that something on it is wrong.
What the entry types mean
- Correction
- A published claim was wrong or overstated. The old claim is superseded and kept, never overwritten.
- Addition
- Something was added that was not there before — a record, a negative finding, a source.
- Release
- A dossier version was published.
Not yet used on this record: retraction, new evidence, clarification, policy. They are listed in the data model because the record will eventually need them, and an entry type invented at the moment it is first needed tends to be shaped by the change it is describing.
2026-08-13
2 entries
- CorrectionCH-004Dossier v0.9
The object header no longer asserts drypoint without saying who said it
The record's headline described the medium as "Drypoint etching". That is the auction house's wording for this sheet, and it was reproduced on the home page, the dossier and the catalogue concordance with no attribution, while the note explaining where it came from sat one page away on the claim. The wording is a minority one. New Hollstein calls the plate an etching and does not use the word drypoint in the entry held here. Of the 26 institutional records in this census, three describe their impression as etching and drypoint; the rest call it an etching. None uses the phrase "drypoint etching". It also sat directly against an open question. RQ-09 asks whether any drypoint burr survives in this impression. A header stating the medium as drypoint answers, in the first line a reader meets, a question the record elsewhere says is untested. The claim itself was correctly sourced and correctly bounded, and is unchanged in substance. What changed: the header now reads "Drypoint etching (Bonhams); etching in most institutional records", the neutral term is what goes to schema.org artMedium and the API, and the claim's stated_by no longer credits New Hollstein with a word it does not use.
Why this changed
A header is read as the record's own position. Reproducing one cataloguer's minority wording without naming them, next to an open question it appears to settle, is an overclaim by placement rather than by sentence.
Affected records
- CL-MEDIUM
- RQ-09
- np-rmb-001
- AdditionCH-005Dossier v0.9
Bonhams declines to add to the 2023 entry; recorded as a negative finding, not as silence
The owner wrote to the auction house that sold this sheet, asking two things: for a view on the state, and whether the consignor would now permit any part of the provenance to be named — offered at four levels, from the family name with dates down to the name of the dealer through whom the sheet was consigned. The department replied the same day that it has nothing further it can share about the lot. That is recorded here rather than left as an unanswered question, because a closed door is a different fact from an unopened one, and because the next person to research this impression should not spend the same enquiry again. Two things are worth stating precisely. The department says it has nothing further it can *share*, which is a confidentiality formulation and not a statement that nothing further exists; the record does not infer which. And the specialist who catalogued the sale — the person who told the owner in 2023 that the sheet came from a German family — has since left the house for a curatorial role. The knowledge of this consignment and the institution that holds the file are now in different places. The undocumented interval from about 1846 to 2023 is unchanged. What has changed is that it is now undocumented for a stated reason.
Why this changed
A negative finding that goes unrecorded gets researched again by the next person. The search path is as much a part of the record as the result.
Affected records
- RQ-11
- RQ-02
- RQ-13
- EV-BONHAMS-2026
- np-rmb-001
2026-08-12
3 entries
- ReleaseCH-001Dossier v0.9
Record compiled and published at version 0.9
NP-RMB-001 receives its identifier and a public record. The compilation drew on the acquisition file — the auction catalogue entry, the payment receipt, the auction house's condition photographs of the sheet, correspondence with the department specialist, and the New Hollstein catalogue entry for no. 170 — together with a survey of institutional collection records for this plate. What that established is on the object record and the state atlas, with the evidence and confidence for each field. What it did not establish is on the research questions board. From this entry onward every change to the canonical record is logged here with its reason and, where one applies, the name of the person who found it.
Why this changed
A record that is not public cannot be corrected by anyone who knows better.
Affected records
- np-rmb-001
- CorrectionCH-002Dossier v0.9
RISD Museum 84.198.1269 corrected from no recorded state to New Hollstein 170, state III
The census row for RISD Museum 84.198.1269 recorded no state, and said in its notes that the museum's IDENTIFICATION, SIGNATURE/INSCRIPTION/MARKS and DIMENSIONS headings were "present but EMPTY". On that basis it also recorded that a web-search summary describing the impression as a third state of four had been checked against the museum record and rejected. All of that was wrong. Those headings are collapsed disclosure panels. Their contents are present in the page as served and appear when the panel is opened. The museum's record reads: STATE iii/iv; STANDARD REFERENCE NUMBER New Hollstein no. 170.iii; Lettered UL: "Rembrandt"; Watermark: bunch of grapes; plate 13.5 x 10.3 cm. The summary that was rejected was correct. The consequence was not only a blank field. This is the only impression in the census catalogued in the same state as NP-RMB-001, and its image is CC0, so the record had been treating its single most useful comparator as an object with nothing to say. It is now published as a comparator alongside the Rijksmuseum impressions of states I, II and IV, and the state III gap in the comparator set is closed. The failure is the same one this record already has on file from its compilation: an absence in what the project could see was read as an absence in the source. There it was a catalogue that was not online. Here it was a panel that had not been clicked.
Why this changed
The project rejected a true statement because it could not see the evidence for it, and then recorded the rejection as a finding.
Affected records
- CN-risd-museum-rhode--84-198-1269
- RQ-02
- RQ-12
- np-rmb-001
- AdditionCH-003Dossier v0.9
All 23 institutional links verified in a browser; a state-count discrepancy opened as RQ-13
Twenty-three institutional URLs in the census and glossary had never been confirmed, because they return 403 or 429 to automated clients. Every one was opened in a browser on 12 August 2026. All 23 resolve; none is dead. Two were canonicalised to the addresses they redirect to (Museum of Fine Arts, Boston, and the National Gallery of Art's legacy object path), and the Canadian Conservation Institute page cited on the glossary for the 50 lux figure was confirmed to carry that figure. Reading the records rather than their status codes turned up something the project had not looked for: the four print rooms cite four different state counts for this plate. The British Museum cites 170.IV, the National Gallery of Art "no. 170, state ii/iii", RISD "iii/iv", and the Morgan gives "NHD 170, II" beside a note reading "Only one state." Each is now recorded verbatim in its census row, and the discrepancy is opened as RQ-13 rather than reconciled.
Why this changed
A link nobody has opened is not a citation, and the records behind these links held more than the links did.
Affected records
- RQ-13
- RQ-02
- np-rmb-001
Changes are logged against the dossier version they belong to, and each version’s full summary sits on publications and data. Superseded claims remain queryable through the public API with their status marked, so a machine reading this record can tell the difference between a claim that was never made and one that was withdrawn.