Dear Simon and Pat,
On Nov 24, 2013 6:27 PM, "Pat Hayes" <phayes@xxxxxxx> wrote:
>
> Quads are not part of official RDF. RDF graphs are (normatively) required to consist of triples. But also, the reification vocabulary is, if not deprecated, widely considered to be archaic and is kept in the spec only for legacy reasons. That section of the new spec is copied almost verbatim from the old one, relegated to an appendix.
I don't think quads can be excluded from [official RDF ], broadly construed, since N-Quads (and I guess TriG) are part of the RDF 1.1 batch of candidate recommendations.
I cannot understand what it would mean for the non SPO part of the quad to be part of the "triple" in the RDF graph (identity criteria alone are too confusing).
I am pretty sure that having the name part of a named graph denote the graph is the least general thing for it to denote, unless a heck of lot of introspection is added.
Treating it as denoting some octet string is a bit too general (though has to happen in some fashion if signatures are being generated
Treating it as denoting a second graph of reified statements that could entail the graph under some extended rules of entailment seems a useful starting point (it handwaves through the introspection).
RDF reification is denigrated to the extent it is not deprecated, but at least quads suggest some ways for not doing it. Provenance for a graph name might be defined so as to entail provenance assertions on reified triples. Queries on the provenance of triples in a graph from multiple sources might usefully be in the form of quad or quad-like things (but must a triple that is in two graphs be two quads, etc.?)
I blame sparql.
[MW>] My choice would be to use the additional element as an identifier for the quad (is that what you mean by reification?) This then means it can be referenced elsewhere to construct named graphs, add meta-data (such as provenance) etc. It’s also essentially how ISO 15926 works.
Regards
Matthew West
Simon