- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Thu, 29 Sep 2011 13:25:29 +0100
- To: rdf-WG <public-rdf-wg@w3.org>
Can we change the title to not imply that graph identification is part of Turtle? Why not a separate langauge using the same langauge? A document of a graph can be read to get a stream of triples; a compound document can't be. Style wise - there is packing a collection of graphs together (@graph emphasises this), writing/reading named grapg data (TriG emphasises this) and dump (N-Quads emphasises this). I hope we have N-quads as standalone spec c.f. Turtle and N-triples. [[ Assumptions: Alignment with SPARQL is more important then alignment with N3. The community does not have a lot of investment in TriG syntax. The community has some investment in N-Quads. ]] While the community may not have a *lot* of investment in TriG syntax it is out there and used. I haven't seen or heard any serious faults with the general design (yes - there are specific issues). To switch to another format needs a positive reason to switch even if the barrier is quite low. Considerations: Should "format X", a syntax based on Turtle used for graphs, 1/ Have Turtle as a subset? i.e. is a valid Turtle document a valid format X? 2/ Have N-Quads as a subset? I used to think (1) was important but I'm not so convinced any more. A document of triples isn't a document about several sets of triples. Possibilities for TriG-ish: A/ TriG as is. B/ TriG as is, but drop the uniqueness of the default graph and naming of graphs (does any system actually impose these conditions? why?). C/ TriG+NQuads Allow N-Quads outside {} D/ TriG+Turtle Allow un-enclosed Turtle for the default graph (as well as or instead of {}) Note that C also means N-triples because N-triples is a subset of N-Quads. These are only general directions - whatever we do there are techncial issues like treatment of trailing DOTs. Andy
Received on Thursday, 29 September 2011 12:26:00 UTC