Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Joshua Westgard volunteered to take minutes.
  • The first item of discussion was Nancy Fallgren and Doron Shalvi's post on the Fedora community list regarding questions of Fedora's URI/URL usage.

  • Doron Shalvi expressed at the outset his recommendation and wish that the conversation would continue at a time when Andrew Woods is available since he is likely to have valuable input on this question.
  • Nancy Fallgren summarized the main points of concern:
    • URIs should be resolvable HTTP URIs.
    • Using URLs rather than URIs is not a best practice (i.e. making statements in RDF that blur the line between web resource and real-world object).
    • IDs for abstract concepts are being minted by various groups and we need to be able to use those in the RDF published through our Fedora instances.
    • Linked data is primarily for machines and not humans.
    • Currently NLM creates identifiers that are referred – referred to a as permalinks to – to represent these real-world objects, but it would be difficult if not impossible to use these directly in Fedora as it is currently defined.
  • Joshua Westgard posited that the central barrier to the sort of use that NLM is describing might be the single subject restriction.
  • Jared Whiklo suggested that one way to talk about real world objects in a Fedora ecosystem would be to use a triplestore alongside Fedora to store RDF about real world objects.
  • Another option would be relaxing the single subject restriction.
  • Aaron Birkland mentioned OAI-ORE's concept of a resource map as another helpful model to consider in this context.
    • This describes a method of keeping object and description separate and yet resolvable
    • Two methods are recommended for dealing with this problem: hash URIs and a the 303 response (see other)
    • Aaron Birkland also expressed support for removing the single subject restriction
  • Danny Bernstein raised the further issue noted in the mailing list thread of the question of the appropriateness of RDF as a backend data storage format, and asked that this question be fleshed out.
  • One issue with RDF is scalability (large datasets tend to balloon to millions and billions of triples rather quickly).
  • Aaron Birkland suggested that the question may really be about whether the web-based architecture is appropriate.
  • Doron Shalvi added two additional concerns:
    • restrictions on URI patterns creates a scalability problem for users with large collections of materials
    • there is a need to make assertions/statements about things external to the repository and single subject makes it necessary to then bury the LDP server under other layers of infrastructure, which in turn calls into question the benefit of using Fedora
  • There seemed to be a consensus that the discussion was valuable and needs to continue, particularly as we move toward finalizing a version of the API specification.

...