Re: RDF-ISSUE-5 (Graph Literals): Should we define Graph Literal datatypes? [RDF Graphs]

Hi Ivan,

On 6 Mar 2011, at 08:28, Ivan Herman wrote:
> Is your opinion that
> - we should clarify the g-* concepts in terms of _concepts_


> - we should not define any extension of any serialization (maybe with the possible exception of N-Quads) 

I believe we should not add multi-graph support to the formats already widely deployed on the Web (RDF/XML, Turtle, RDFa, and maybe N-Triples) because that would have security implications and would lead to major headaches for anyone who currently uses named graphs or an equivalent mechanism to manage content off the Web.

Regarding new formats, I'd say do as few as necessary. Prefer simple ones. Prefer ones that are already in use. The simplest option would be “do nothing”. The next simplest one would be “only do a REC for N-Quads”, and personally I'd be very happy if we could get away with this. At any rate, I hope that the collected use cases would be helpful in discussing the necessity of these formats:

> - declare victory

Oh yes, definitely!

[I moved the following bit down here as it's an aside to the discussion above]
> I have the impression that many (if not all) of your arguments would also apply to TriG, ie, they are bound to graph literals. 

TriG wouldn't require graph literals. It could be defined as serializing an RDF Dataset as currently specified in SPARQL, that is, a collection of (URI, g-snap) pairs plus a default g-snap.


Received on Sunday, 6 March 2011 12:08:50 UTC