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
Attendees
- Danny Bernstein
- Peter Eichman
- Bethany Seeger
- Jared Whiklo
- David Wilcox
- James Silas Creel
- Andrew Woods
- Kevin Ford
- Aaron Birkland
Agenda
Accessing Modeshape layer WebACFilter :
Fedora 5.0.0 Preparation Tasks
Work completed in the last week
- CTS - 2 tests written
- Review State of the API Alignment Effort
- Outstanding PRs
- Review the Jira List
- Documentation
- In Review:
- Status of key Jiras:
Ideas for this week
- Low hanging fruit (once 2742 is merged)
- Do a code review
- Join the CTS Challenge: Write two tests before the next meeting:
- Look at how easy it is to add tests: https://github.com/fcrepo/Fedora-API-Test-Suite/commit/18a58618a856c6796fed69765d796985048f345b
- A little higher in the tree:
- Low hanging fruit (once 2742 is merged)
September Sprint Planning
Ticket Summaries
Please squash a bug!
Tickets resolved this week:
Tickets created this week:
Minutes
- Brief discussion about Amherst College's acrepo apache camel services: https://gitlab.amherst.edu/acdc/repository-extension-services
- perhaps more conversation about what these are at a later meeting.
- Accessing Modeshape layer WebACFilter: issue
- adding Acl:append checking - there are some difference in how Acl:append works in POST: for example you can post to a RDF source but not a non-RDF source. In looking up the nodes in modeshape, via spring configuration, things are not actually getting injected into the webac filter. Peter may have been using the wrong configuration file to do this with.
- webac code is in the WebACFilter - but perhaps some of the checks belong in the FedoraLdp class, so that access to node data is readily available. Concern with doing that is that authorization would be split between to two places and do we want authorization that close to the HTTP API core?
- webac is like a shield around the HTTP interface. Maybe a way to make a call that's unfiltered? We need to be able to distinguish requests that need to be filtered and ones that do not (internal requests), to ensure they are not going through the authorization chain again. One solution could be internal admin credentials?
- Current approach is to pull in modeshape (fedora repo impl) to resolve question of what type of RDF resource it is. That resource is not currently being injected into the webac. Possible not configuring the correct spring file.
- What about cases where authorization is not just happening out front - ie. cases like recursive delete. That's an area where authorization checks might need to happen outside/after of the WebACFilter.
- Jira ticket for looking at WebAC and recursive deletes where authorization might be different for child nodes.
- one possibility is that a recursive delete will delete what user is authorized to delete – or reject the request outright.
- Andrew pointed out that moving the WebACFilter stuff into the HTTP layer is more out of necessity rather than architectural design –
- more sound/maintainable approach by having the filter do it all, but barrier is getting the injection of fedora objects to work correctly
- generating more HTTP traffic when we're already inside the server isn't ideal
- Jared brought up the wizardOfOz approach (hidden user) - an internally credentialed (hidden) user whose requests are always just accepted by server. Perhaps tie it to the servers IP, so they are only locally generated requests.
- Short term plan - move some WebACFilter stuff closer to the HTTP layer. Danny will look at this some. Fresh eyes on this welcome.
- Fedora 5.0.0 Preparation Tasks
- CTS – all the tests are stubbed out now.
- Documentation changes have slowed down. Still work to do there.
- Outstanding PR's awaiting review from folks - just one current one: https://github.com/fcrepo4/fcrepo4/pull/1386
- some JIRA tickets just need to be closed (the PR's were merged already)
- JIRA tickets:
- Bethany will look at and wrap it up this week.
- There are just 10 issues left for the 5.0 API alignment.
- Looked at
- Does it still apply? How do we want to approach this? This was based on a use case where some resources ought to be public and not require authorization. Aaron B. thinks it's still pertinent. This was to see if the refactor for Shiro changes this behavior. This question is still open.
- There is code in the WebACFilter that supports authorizing unauthenticated requests. The bigger question is how to send unauthenticated requests - right now that's being handled by something external to the web app – ie front end talking to shib or the handled by the servlet container. It's either everything is authenticated or nothing is. Having Shiro might open up options in the future, for authentication, with the ability to configure Shiro authentication realms (a potential long term solution).
- Fedora will do the right thing when an unauthenticated user makes a request - acl is still checked.
- Aaron B. will re-scope it to documentation: documenting where the repository stands on working with this feature.
- How are folks tying into things like shibboleth - Aaron's group is using a proxy-ized path.
- LTS update: the policy has been voted in by Fedora Leadership and is now in effect.
Actions
- Danny to explore inserting a macro into documentation for Fedora version.
- Send email to community seeing who is interested in reviving OAI
- Aaron B. will re-scope to be a documentation issue:
- Bethany will look at this ticket and wrap it up this week:
Notes
