- From: Ian Davis <ian.davis@talis.com>
- Date: Thu, 21 Jul 2011 01:08:18 +0100
- To: RDF WG <public-rdf-wg@w3.org>
I think re-introducing the word "graph" into these new terms perpetuates the confusion led to the need for g-* terms in the first place. We recognise that "graph" has subtly different semantics between sparql and rdf concepts so let's avoid that term. Here's my suggestion, which I think are unambiguous: g-snap: "(mathematical) set of triples" g-box: "container of a set of triples" g-text: "serialization of a set of triples" One step further could lead us to coin a new term: TripleSet g-snap: "TripleSet" g-box: "TripleSet Container" g-text: "TripleSet Serialization" A TripleSet is immutable. A TripleSet Container contains exactly one TripleSet at a time but could be a different TripleSet at different times so a TripleSet Container is mutable. A TripleSet Serialization serializes exactly one TripleSet. A quick Google search suggests TripleSet is not a term in common use for other systems. In terms of spec changes: replace every occurrence of RDF Graph in the RDF specs with the term TripleSet I think it would be useful to talk about some of the characteristics of these concepts e.g. equivalence Two TripleSets are equivalent if they conform to the bijection defined at http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/#section-graph-equality (i.e. they differ only in the identity of their blank nodes). Two TripleSet Containers are equivalent if their contained TripleSets are equivalent Two TripleSet Serializations are equivalent if they parse to equivalent TripleSets In terms of those "terrible TAG/REST terms": A URI can denote a TripleSet Container. Dereferencing that URI should return a representation consisting of the TripleSet Serialization for the TripleSet currently contained by the TripleSet Container. A user agent parses the representation to derive the TripleSet which they will most likely place into a local TripleSet Container. In terms of SPARQL, a dataset consists of TripleSet Containers: ( C, ( Ui, Ci ) ) A more concise name for TripleSet Containers would be a nice to have. Talis has been using the term Metabox for this concept for a long time (no prior art, I only recognise the equivalence today :). I don't think that's a great term to use, but perhaps TripleBox might work? Now, sorry to do this to you all, but I am away on holiday after tomorrow so I won't be around to get into any discussion this email may generate. I weighed up whether to send it now or wait and decided it was best to get something sent earlier. I'll pick up any conversation in a couple of weeks. Ian -- Ian Davis, Chief Technology Officer, Talis Group Ltd. http://www.talis.com/ | Registered in England and Wales as 5382297
Received on Thursday, 21 July 2011 00:08:51 UTC