Time/Place

This meeting is a hybrid teleconference and IRC chat. Anyone is welcome to join...here's the info:

Attendees

Agenda

  1. C4L report
  2. Identifiers: paths and hierarchy vs. opaque ids and object buckets
  3. RDA report
  4. Authorization approaches
  5. other...

Minutes

 

awoods: Sprint is over. People were at code4lib and RDA. Not much got done.

code4lib report:

esme: Had a good time. Lots of repository discussion, mostly Hydra-related. Fedora 4 talk went well. Tried to get people excited about getting engaged. A few people were surprised that Fedora is all-volunteer. Avalon had a presentation. Some people were discussing RDF support in Hydra, including support for Fedora 4. Not clear that any real progress is being made.

awoods: jcoyne seems to have set up a Travis build of Fedora 4 with a Ruby library going against it. cheer and jcoyne will likely be putting in effort as OR approaches.

esme: cheer and jcoyne are working on Spotlight, a Blacklight plugin.

gregjansen: UNC library is nice.

Hierarchy and performance and identifier translation:

awoods: Fedora 3 is a big bucket of objects, with relationships. In Fedora 4, you can do that, or you can use hierarchy supported by the JCR. But having too many objects in the same directory (maybe 1000-10,000) performance degrades. We will be working with MODE to fix this.

We are interested in separating the internal and external identifiers for resources. ajs6f is working on making the translation between them more sophisticated.

People are likely to start using Fedora 4 the same way they use Fedora 3, as a big bucket. They will run into the performance problem. We may need to support the mapping of a "flat" space of identifiers into a hierarchical space that takes advantage of MODE's performance characteristics, as part of our subsystem for identifier translation.

kevinclark: Could you use a pair-tree mapping?

ajs6f: Sure.

gregjansen: What are hierarchical identifiers made from?

ajs6f: UUIDs.

awoods: You could use a hash of the "incoming" identifier.

gregjansen: No. Then you'd have to move content around if the identifier changes.

awoods: Also, more education is needed, especially for legacy users.

esme: That is not surprising. It will take people a while to get this message.

awoods: Yes.

awoods: Also, we can test with different backstore configurations to see whether this problem is ameliorated by using something other than LevelDB.

https://www.pivotaltracker.com/story/show/68662964

benpennel: We have seen bad performance degradation with MODE in some configurations.

RDA report:

awoods: David Wilcox and awoods were at RDA. Encouraging the use of repositories in the research community was of high concern. More than 400 attendees. At RDF there are birds-of-a-feather groups, which turn into interest groups, which turn into working groups. There was a Fedora birds-of-a-feather with about 25 people. It will probably turn into an interest group. It's a way for Fedora to be "in touch" with the "research community". There was also a Fedora steering group meeting.

davidwilcox: There was good attendance at the birds-of-a-feather session. davidwilcox is trying to follow up with attendees. He will be creating something about how Fedora supports research work.

Authorization:

There is a basic role-based access control framework in place, but much more needs to be done. There are use cases for performing authorization based on:

• the presence or values of selected properties
• the presence of mixins
• IP address
• indirected policies stored in the repository

gregjansen: We've only tried to address IP so far. There are extension points for this stuff. But maybe that's not the best answer.

awoods: Yes.

gregjansen: With shared hierarchy, policies could be shared more widely.

awoods: Is XACML the best?

gregjansen: Potentially.

mikedaines: XACML is hard to write. We don't want to have to write XACML ourselves.

gregjansen: We've been burned by XACML in the past.

awoods: Can XACML be used for these use cases?

gregjansen: We can put mixins and things into the XACML context. But only if the original user is addressing Fedora directly.

awoods: That's the responsibility of the client.

esme: Apache uses some header to include the original client IP address.

awoods: Is it possible?

gregjansen: It's possible.

awoods: It would be great if we could think through this outside of sprints.

gregjansen: We can make a strawman at UNC.

awoods: We can have a special topics call about this.

gregjansen: We'll make a strawman. And we'll take into account the stuff Eric James uncovered about federated content.

Stefano Cossu: We're fixing https://www.pivotaltracker.com/story/show/64929156, using basic authentication.

awoods: That's great. You want to use mixins for auto-generated properties.

https://wiki.duraspace.org/display/FF/Art+Institute+of+Chicago+Use+Case+-+Structural+Validation

stefano: Yeah, but that might not be a problem for the repository. I'd just like to know which auto-generated properties are related to a given mixin. That is not exposed in the RDFS endpoints for types.

awoods: stefano is doing things that other Fedora users aren't.

stefano: Yes.

awoods: Let's stay in touch to make sure we put our limited resources to good use.

 

 

Actions

1 Comment

  1. Stefano Cossu

    For the issue with auto-generated properties and child nodes, see Re: Art Institute of Chicago Use Case - Structural Validation. 

All content on the LYRASIS Wiki is licensed under the CC BY (Attribution) license, unless otherwise noted.