- From: Danny Ayers <danny.ayers@gmail.com>
- Date: Tue, 11 Jul 2006 20:02:34 +0200
- To: "Mark Baker" <distobj@acm.org>
- Cc: "Henry Story" <henry.story@bblfish.net>, www-archive@w3.org
On 7/11/06, Mark Baker <distobj@acm.org> wrote: > If the sender of the message intends the message to communicate > inferred triples, then that needs to be indicated in the message > itself and an inference-specific media type would be appropriate. > Otherwise they're not communicating it because nothing in the RDF/XML > spec licenses inference meaning that existing RDF/XML processors > wouldn't extract the inferred triples from the message. > > It's all about intent. See the example here; > > http://lists.w3.org/Archives/Public/www-rdf-comments/2003JulSep/0342.html > > FWIW, the Sem Web specs should have either; > > a) minted a new media type that assumed inference, or > b) explained that inference isn't implied by the rdf+xml media type, > and therefore that inferred triples need to be explicitly included in > the RDF/XML document if the message sender intends to communicate > them. > > Unfortunately they did neither. Once RDFS comes to be commonly > exchanged, this problem will be discovered and the easiest fix will > likely be a new media type. While I'm not really convinced by Henry's arguments against APP getting new mime types, I'm not really convinced the RDF/RDFS/OWL situation is such a problem. Ok, by using the application/rdf+xml the publisher is only providing data according to the base RDF model. I'd suggest that any decision about what can be inferred is up to the consumer, I can't see how it could be feasible otherwise - how does application/rdf+xml+owl-lite+n3-rules sound? What if the info is OWL DL, yet has been accumulated at the client alongside OWL Full data - is the consumer licensed to apply OWL Full inference, or should the whole lot be rejected as inconsistent? To me this seems analogous to the case of publishing HTML: <h1> is the most important heading, but it's up to the client agent what to do with it. If the publisher's intention is for it to be rendered as 72point Courier, then that can be stated in the CSS. Similarly, if the RDF publisher wants the consumer to accept the inferred statements as asserted facts, then it's up to the publisher to do the inference prior to publication. Cheers, Danny. -- http://dannyayers.com
Received on Tuesday, 11 July 2006 18:02:50 UTC