- From: Richard Cyganiak <richard@cyganiak.de>
- Date: Wed, 29 Jun 2011 00:03:01 +0100
- To: David McNeil <dmcneil@revelytix.com>
- Cc: RDB2RDF WG <public-rdb2rdf-wg@w3.org>
On 28 Jun 2011, at 20:11, David McNeil wrote: > Does adding this help? > > [[ > Conforming R2RML processors MAY accept R2RML mapping graphs encoded in other RDF syntaxes besides Turtle. > ]] > > Richard - I think that would match what we discussed today on the working group telecon. Others please speak up if this is not the case. > > (Perhaps I am over-analyzing it but when I read that statement I imagine the following scenario: some implementation supports RDF-XML. They are within the spec. Does this mean that a user could create a mapping for this implementation, represent it with RDF-XML and still be within the spec? No. It's the right graph, but the wrong RDF syntax. It would be a conforming "R2RML mapping graph", but not a conforming "R2RML mapping document". An "R2RML processor" MAY accept it, but is not required to do so. I'd say that ideally, the R2RML spec should clearly specify three things: "R2RML mapping documents" -- a class of documents, Turtle-based "R2RML mapping graphs" -- a class of RDF graphs, could be serialized as anything "R2RML processor" -- a system whose input is an R2RML mapping document and an RDB, and whose output is a (possibly virtual) RDF dataset. Along with conformance criteria for each of the three. Must crucially, a system must pass all test cases and implement all MUST requirements to be considered a conforming R2RML processor. A mapping graph must fulfil certain syntactic requirements. A mapping document must serialize a conforming mapping graph. I'm not sure how close we can get to this ideal picture, but it's what I'd like to achieve if we have sufficient time and input for the test cases. Best, Richard
Received on Tuesday, 28 June 2011 23:03:30 UTC