Time/Place
This meeting is a hybrid teleconference and slack chat. Anyone is welcome to join...here's the info:
- Time: 11:00am Eastern Daylight Time US (UTC-4)
- Audio/Video Conference Link: https://duraspace.zoom.us/my/fedora
Dial-in:
+1 408 638 0968
+1 646 876 9923
+1 669 900 6833
Meeting ID:
812 835 3771
Join fedora-project.slack.com on the "tech" channel
Attendee
- Danny Bernstein
- Jared Whiklo
- David Wilcox
- Andrew Woods
- Ben Pennell
- Peter Eichman
- Jon Roby
- Bethany Seeger
- Aaron Birkland
Agenda
- Announcements
- 2019-05 Fedora Camp Atlanta
- OR Events
- FMP Grant Survey
- Updates from Slack conversations
- #tech
- #ocfl
- CNI / Leaders / Grant / ACRL Debrief
- Ready for Review:
- Import / Export
- Next Steps for Fedora 6
- Java 11?
- <your discussion point here>
In Review
Please squash a bug!
Tickets resolved this week:
Tickets created this week:
Notes
Announcements
- Fedora Camp Atlanta
- May 20-22 at Emory
- Andrew, Danny, Jared as instructors
- Open Repositories events
- Fedora user group on Friday
- Dinner earlier in the week
- FMP Grant survey
- Survey went out yesterday
- Open to anyone to complete. About 10 minutes.
Updates from Slack
- #tech
- Some questions about using JSON-LD
- Questions around scale, ingest performance, etc.
- #ocfl
- Conversation around locking
- Some suggestion that implementation notes should mention different approaches to contention in object updates, uncoordinated applications adding revisions to the same object. Maybe out of scope?
- Some commenters thinking of scenarios analogous to those we will deal with in Fedora (e.g. many writers making simultaneous changes to an object).
- How to build up and commit changes to an OCFL object safely. This could go in the implementation notes.
- Dealing with locking: Fedora will need to answer this question. Less about OCFL and more about the applications built on top of OCFL.
- Fedora design includes caching changes until committing to an OCFL object.
- Changes are maintained in a deposit space until a version is explicitly created.
- OCFL in S3: filesystems can make certain guarantees that are impossible in S3. File paths are not really directories, no true renaming. S3 may only be safe if every client uses a single application that mediates access.
- Can locking be cast a a general HTTP issue rather than a Fedora-specific issue?
- Stateless HTTP requests will need to be managed by a higher-level application
- It would be useful to have concrete use cases
- Oxford experimenting with Aaron Birkland's OCFL client
- Created 250K objects and were pleased with performance
- Conversation around locking
Ready for Review
- Most of the way through this issue
- Test still failing in Travis but not locally
- Null pointer exception, not clear where this is occurring
- Andrew will try a couple builds and report back
CNI / Leaders / Grant / ACRL Debrief
- Fedora Leaders meeting
- We have consensus for moving forward with the Fedora 6.0 design plan
- OCFL instead of ModeShape, implementing the Fedora API, adding a query API
- Strong interest in not releasing Fedora 6.0 without migration support
- Some interest in funding testing.
- Re-centering on the possibility of swappable back-ends
- Some people may not want OCFL, concerned with optimizing performance and assuming OCFL would have some negative impact
- Vote coming soon on release of Fedora API spec
- We don't yet have 2 full implementations but this isn't a hard requirement in the charter
- Andrew and David will reach out to current implementers to assess status
- We have consensus for moving forward with the Fedora 6.0 design plan
Actions
- Peter Eichman is planning to work on documenting UMD's fixity check system on the Fedora in Production: Case Studies wiki page.
- Danny Bernstein will pull together some use cases for object locking in OCFL
