triple store interest group 2016.01.15
my own notes from the first meeting.
aaron coburn - amherst - sidecar triplestore
nabeeld - michigan - research data
trey - princeton - active fedora developer - has done extensive linked data work in the past
tom johnson - dpla - they use some hydra tooling with a triplestore already. nominally the maintainer of active triples and other related projects: rdf-ldp libraries, blazegraph repository ruby interfaces, rdf.rb 2.0 (adding an ACID transaction interface)
Mark matienzo - dpla - also has concerns around implementation w/r/t hydra-in-a-box. using active triples and LDP to manage metadata
hector - penn state - interested in what others are doing.
Adam wead - penn state - wanting to learn more.
penn at cornell - learning.
matt - ucsd - have had a variety of triplestore implementations. currently use a postgres triplestore backend but migrating to f4 and thinking about a sidecar triplestore as well. get a lot of request from metadata specials for running queries, metadata cleanup opportunities, etc. going to run a sparql workshop for their metadata folks.
christina - metadata workflows
justin - dce - lurking to see what everyone’s interested in doing
sheila - also lurking.
sidecar triplestore vs. linked data fragments
triple data fragments give sub, pred, object and return triples
goal: instead of that have a web interface where you give it a subject and it returns cached triples, then use an http API.
multiple implementations: one uses marmotta, one uses general RDF
2nd phase (yet to be done): sidecar indexer for atomic updates of linked data in SOLR.
hydra people talk more about importing linked data
fedora groups talk more about exporting triples from fedora to a triplestore. Use cases:
- inferencing
- using sparql for analytics - ad hoc queries across the data set in ways not possible w/ fedora or sold
- validating integrity of triples
RDF Shapes working group has produced some validation language. fedora’s API-X group is also working on validations.
Access controls over a triplestore? Hydra does this well by controlling access to solr.
Tom J:
- build community knowledge about which triple stores have various capabilities. get a sense of what people are able to use / who might maintain the various projects.
- gap analysis / development work on plugging this into the stack.
blazegraph: tom j. wrote the gem. it’s performant and good but if you use blank nodes they will duplicate!! user beware. also potentially trouble around quads.
marmotta
jena / fuseki at scale (aaron coburn)
virtuoso - and attendant unmaintained gem.
apache rya