Skip to content

RFC 090: record the stub requirements for digitisation - #175

Open
kenoir wants to merge 1 commit into
mainfrom
rk/rfc-090-digitisation-stubs
Open

kenoir wants to merge 1 commit into
mainfrom
rk/rfc-090-digitisation-stubs

Conversation

@kenoir

@kenoir kenoir commented Oct 13, 2026

Copy link
Copy Markdown
Contributor

Preview

https://github.com/wellcomecollection/docs/blob/rk/rfc-090-digitisation-stubs/rfcs/090-axiell-folio-sync/README.md

What does this change?

RFC 091 makes the FOLIO instance HRID the preservation identifier after migration, and wellcomecollection/platform#6697 put visual and AV material in Axiell, synced to FOLIO. The stubs this sync creates are what the digitisation export resolves, so the RFC now says what they carry.

  • The instance HRID is assigned by FOLIO and the instance is found again through a GUID mapping table; holdings and items keep their computed keys.
  • A new section covers the identifier set, instance type per category, the stub marker and discovery suppression, checked against catalogue-pipeline main and the hosted dev tenant on 2026-10-13.
  • Open questions gain the Collections Information decisions, the archivecatalogue restriction on the Axiell feed, and the dev tenant mismatch.

How to test

.scripts/validate_rfc.py and .scripts/create_table_summary.py --check-readme pass; the preview renders the new tables.

How can we measure success?

The sync change can be built from this RFC and the spec in location-movement-control-docs without reopening the identifier decision recorded in wellcomecollection/platform#6669.

Have we considered potential risks?

The 035 and 024 emission for stubs is inferred from mod-oai-pmh's rules rather than observed; the RFC says so and makes observing it an acceptance step.

The instance HRID is now assigned by FOLIO from the shared in sequence
rather than computed as AxC-instance-{guid}, because RFC 091 makes the
HRID the preservation identifier for anything digitised after the
migration. The instance is found again through a GUID to instance
mapping table with a self-heal query on the stored bare GUID; holdings
and items keep their computed hrids.

A new section records what a stub has to carry for digital production's
export tool: the identifier set (GUID twice, object number, RefNo under
its own type, predecessor b-number from 035 Bibliographic Number or a
b-number-shaped 907), an instance type per category, the stub marker
that replaces the AxC- prefix, and discovery suppression on create.

Open questions gain the Collections Information decisions these need,
the archivecatalogue restriction on the Axiell feed verified on
production on 2026-10-13, and the dev tenant mismatch.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant