- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Tue, 19 Jun 2012 11:10:24 +0100
- To: public-rdf-dawg@w3.org
On 19/06/12 02:50, Gregory Williams wrote: > I ran across an issue I was hoping for some input on. As a result of > working on the protocol tests, and reading David Booth's recent > comment on how datasets get constructed[1], I looked over the > documentation for sd:DereferencesURIs[2] which can provide some > metadata that may help David's issue (and for which I've got an > action to respond to his comment). > > The motivation for sd:DereferencesURIs was that some implementations > (including mine) were already doing dereferencing to construct a > dataset for query, and I just extended that in the spec to cover > update, too. But now that I think more closely about it, I'm not sure > it makes much sense. The current SD text reads: > > """ sd:DereferencesURIs, when used as the object of the sd:feature > property, indicates that a SPARQL service willdereference [AWWW] URIs > used in FROM/FROM NAMED and USING/USING NAMED clauses and use the > resulting RDF in the dataset during query evaluation. """ > > I'm worried that the "USING/USING NAMED" part of that doesn't make > any sense. That is, if an implementation constructed a dataset on the > fly based on a USING clause, what use is there in performing an > update operation? Presumably next time an update is submitted with a > USING clause, the updated content in some underlying graphstore may > again be masked by the dereferencing feature. In a single update request there may be operations that use USING/USING NAMED and some that do not. So one operation may use USING/USING NAMED to get a dynamically constructed dataset and the next operation does not use USING* so it can see the results of an earlier operation. Andy > Does anybody have any thoughts on what the best way forward on this > is? I can imagine this could get through REC without anyone noticing, > but I'm not sure I'd be happy with that. Do you think it's worth the > effort to make the change before moving to CR? I'm not aware of any > implementations that do dereferencing for update (though there may be > some), but I suspect a change like this would require another LC(?). > Since we're already doing another LC round for Query, would this be a > big deal? > > Interested in any thoughts you might have. > > thanks, .greg > > > [1] > http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2012Jun/0002.html > > [2] http://www.w3.org/TR/sparql11-service-description/#sd-dereferencesuris >
Received on Tuesday, 19 June 2012 10:11:48 UTC