Joseph, see inline... On Jun 10, 2010, at 5:12 PM, Joseph Boyle wrote:
> This has some good discussion and links both in the post and comments: > http://decentralyze.com/2010/03/09/rdf-meets-nosql/ Very interesting, thanks. OTOH I'm learning less and less new information as I read along in this area. We're approaching the state-of-the-art here (sad to say). > > One point made is that SPARQL doesn't express the full range of queries you > can do in a graph db. Has it been a good match for the kind of queries we do > in Higgins? We don't use SPARQL at all. Higgins has a Java EAV (entity attribute value) graph model interface called IdAS (a so-called "node-centric API (as opposed to a triple-centric API) in RDF-speak"). IdAS can currently only be exposed on the net using XDI (and thereby updated/edited). Higgins would have to adapt a SPARQL engine to sit on top of the IdAS API. Some engines use a standard SAIL API to access the graph data, so I thought we could build a SAIL to IdAS adapter, but I've studied SAIL and there's too much of a mismatch. One more thing, SPARQL 1.1 will be out soon and it will support read/write operations. Until then it is useless in its 100% standardized form. > > On Jun 10, 2010, at 8:38 AM, Sergey Lyakhov wrote: > >> Joseph, >> >>> Are you saying that integrating any graph database will take the same 2-3 >>> weeks of effort? >>> >>> - NG4J >>> - Other RDF quad stores >> Suppose yes. Also, I have to note that there is a modification of SPARQL to >> query quards (we use SPARQL to query data from CP). Suppose, there possible >> some incompatibility problems between triple- and quard-based CPs. >> >>> - Graph databases not specialized in RDF, like Neo4j, InfoGrid, etc. >> Suppose yes for Neo4j, AllegroGraph, Boca. What about InfoGrid, HyperGraphDB >> - as I understand, they does not support RDFS/SPARQL. As a result, for these >> DBs we can implement CP with no query support. >> >>> - Higgins XDI endpoint >> There already exists some (perhaps outdated) IdAS CP implementation used XDI >> endpoint. >> >>> Also if we integrate two or more - would that be like 2-3 weeks each, or >>> some smaller increment like 1-2 weeks per >>> additional database provider? >> >> Yes, suppose we need to implement base CP with data providers support. >> Actually, the current Jena CP can be easily refactored in such a way. >> >> Thanks, >> Sergey Lyakhov >> >> On Tue, 8 Jun 2010 10:49:01 -0700 >> Joseph Boyle <[email protected]> wrote: >> >>> >>> On Jun 8, 2010, at 10:34 AM, Sergey Lyakhov wrote: >>>> Actually, I suppose it should take about 2-3 weeks to implement proposed >>>> NG4J CP. But I suppse we need to harmonize >>>> all architectural questions (authentication, access control, schema/model >>>> context, metadata, etc) before. Also, my >>>> comments are related to the whole idea about using named graph >>>> (subcontext) for metadata on attribute. Not only >>>> NG4J can be used to implement it. Neo4J, Infogrid, etc., can be used in >>>> the same way too. >>> >>> Sergey, >>> >>> Are you saying that integrating any graph database will take the same 2-3 >>> weeks of effort? >>> >>> - NG4J >>> - Other RDF quad stores >>> - Graph databases not specialized in RDF, like Neo4j, InfoGrid, etc. >>> - Higgins XDI endpoint >>> >>> Also if we integrate two or more - would that be like 2-3 weeks each, or >>> some smaller increment like 1-2 weeks per >>> additional database provider? >>> >>> Thanks, Joseph_______________________________________________ >>> higgins-dev mailing list >>> [email protected] >>> https://dev.eclipse.org/mailman/listinfo/higgins-dev >>> >> >> > > _______________________________________________ > higgins-dev mailing list > [email protected] > https://dev.eclipse.org/mailman/listinfo/higgins-dev _______________________________________________ higgins-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/higgins-dev
