Applied Linked Data Call 2015-05-28

Owned by Bill Branan
Jun 16, 2017

Attendees:
 

  • Steven Anderson (Boston Public Library)
  • Trey Terrell (Oregon State University)
  • Tom Johnson (DPLA)
  • Arwen Hutt (UC San Diego)
  • Aaron Coburn (Amherst College)
  • James Van Mil  (University of Cincinnati)
  • Corey Harper (New York University)


Regrets:

  • Mark Matienzo (DPLA)
 

Linked Data Fragments Sprint Scope + Planning Discussion.

  • Search Fragments draft: https://stackedit.io/viewer#!provider=gist&gistId=a226bd32fdc9da86ea55&filename=searchFragment.md
  • Spinning up something that speaks Hydra Ontologies.
    • Start with "Give me a subject, I'll give you triples"
      • This is a sub-type of Triple Pattern Fragments where _subject_ is in the request: { subject ?p ?o . }
    • Works against remote triple store (ie id.loc, getty, etc)
  • Deliverables: 
    • A Hydra Ontology endpoint that caches resources from remote dataset and returns: { <subject> ?predicate ?object . }
      • Q: How does this fit in with QA?
    • Question on when something isn't cached, does it indicate it will be available and then try to cache it? Or block until it is fetched to be cached?
    • Configure some time of cache time limit on a result. Reply with when it was cached as well?
      • Mention of Varnish?
        • But want to design something that is independent of any caching implementation.
    • Question about blocking. If not cached yet, block until cached, or say not cached come back later?
      • Not blocking, redirecting could be interesting for asynchronous use cases
        • Ticket 1: Blocking support first.
    • Desire to add Linked Data Fragments stuff to rdf.rb where possible (core set of rdf libraries for ruby - http://ruby-rdf.github.io/).
      • Do this work in ActiveTriples namespace.
      • Unlicense (http://unlicense.org/) license under the ActiveTriples github to avoid CLA and other license issues.
    • Start of Sprint is June 8th.
      • For those at Open Repositories (Steven, James, Aaron, and Mark), start an email chain to work out schedules.

 

Broader / Narrower SKOS concepts in an application.

  • Given "Sports", get narrower like "Baseball". Given "Baseball", get the broader of "Sports". All in the same response.
    • How to weight this?
    • Will it be confusing to the user? If they search for "Sports", they may really want books just about "Sports".
    • Seems that Broader likely doesn't have a good use case. Like "North America" giving you "Earth" isn't useful.
  • Trey is going to experiment with implementing this over the next week or two.
    • Report back on how that went on the next call. No user testing really but can talk about the implentation experience.

 

Continued "Sidecar triplestore" best practice / implementation discussion.

  • Fedora Replication System ( https://github.com/fcrepo4-labs/fcrepo-camel-toolbox/tree/master/fcrepo-indexing-triplestore ) from fcrepo-camel-toolbox suggested over the polling.
    • But believe this is only limited to when a Fedora object changes rather than the case of just a label being out of date normally.
  • Abstracting out idea of an "event stream"
    • Tells system to "reach past the cache"
  • Need to support HEAD requests on Linked Data Fragments (use case of "update the cache but I don't care what you return").

 

Next Meeting:

  • Due to Open Repositories, the next meeting will be June 25th at 9:00 AM PST.