...
Agenda
- Announcements
- Check-in regarding the double meeting format
- Status of https://github.com/pwinckles/ocfl-java-parent move into ocfl Github repo
- Opportunities to chip in:
- API Test Suite PRs
- Minimal 4 →5 migration needs testing and code review:
- API Test Suite PRs
- Update on Fedora 6 Pilots
- Sprint Planning
- 6.0 Architecture Review
- Versioning review from last week:
- Clarification of the proposal:
- OCFL transactions always result in new versions; we will not support unversioned content in Fedora
- From Fedora's point of view, the current state of a resource is the most recent version (ie HEAD)
- By default Fedora will display only "significant" versions in the list of mementos.
- "Significant versions" are OCFL objects that contain a marker file in the content directory (possibly something like content/.fcrepo/memento)
- Implication: versions cannot be removed ( because removing content from OCFL is likely to be controversial). Copy/Delete entire OCFL object would be the only way to remove version history.
- Questions:
- Is it important to be able to have Memento timestamps synchronized across a multi-object transaction? In other words, are users going to want to be able to version changes across a single time-slice?
Ie:
http://localhost:8080/rest/object1/fcr:versions/20190822122001
http://localhost:8080/rest/object2/fcr:versions/20190822122001
http://localhost:8080/rest/object3/fcr:versions/20190822122001
- Is it important to be able to have Memento timestamps synchronized across a multi-object transaction? In other words, are users going to want to be able to version changes across a single time-slice?
- Clarification of the proposal:
- Multi-object transaction implementation ideas
- Versioning review from last week:
- Transaction Sidecar Spec Update
- 6.0 Architecture Review
- Status on organizing a Fedora documentation review
- Your topic here...
...
Paul Cummins raising the concern about bloat with respect automated versions. His Fedora 3 application performs frequent updates to binary content. In a world of version on change, he would quickly experience bloat. "Copy into new OCFL object + Delete old OCFL object" mechanism for removing unwanted object history is an acceptable solution.
There was consensus shared on the understanding of the new versioning proposal and its implications.
PART 2
Ben Pennell , Peter Winckles and Danny Bernstein continued the discussion of the new versioning proposal.
Everyone seemed concerned that this would represent a significant departure from the current interaction model of Fedora.
Also storing "significant version" files in OCFL seems a bit convoluted. In order to restore a previous version, clients would need to indicate whether or not to make the restored version "significant". Fedora or possibly the client would then need to remove or keep that marker file accordingly.
Question: What is the standard for rebuildability from OCFL? Are we still in agreement that you should be able to take an OCFL layout, copy it to a new location, stand up a new fedora instance against it, and have the state of Fedora be exactly what it was before the move? Can any non-derivative state live outside OCFL?
Actions
- Aaron Birkland to look explore notion of OCFL client with database as authoritative metadata source + asynchronous writing of the inventory.json file
- Peter Eichman and maybe Ben Pennell to make recommendations re transaction side car specification.
- Andrew Woods will look into java 11 transition
...