Re: Hidden triples and self-description

Mark Baker wrote:

> Oops, sorry, should have said "those that know the *RDFS* rules of
> entailment".  Follow?  

Yep :)

> Hence my RDFS media type suggestion, as it
> communicates that there are different entailments at work (i.e. makes
> the message more self-descriptive).

I like this idea, but... when issuing new mimetypes, we need to 
establish whether the mimetype is meant to indicate a /semantic/ 
extension to RDF (RDFS, OWL, n3) or a syntactic/serialization option 
for shipping triples around - for example

   application/rdf+ntriples
   application/rdf+xml

is the same semantic stuff (graph+entailments), but

   application/rdf+xml
   application/rdf+owl

is not. And these:

   application/rdf+owl-full
   application/rdf+owl-lite

are possibly not (apologies, are there mimetypes for OWL yet?).

The key thing is that calling something a 'semantic extension to 
RDF' means its not RDF anymore.

It all potentially gets v. messy when you start mixing and matching 
the various models sitting above RDF within a single entity. An 
Lbase mimetype perhaps makes things more confusing again. I don't 
even know whether mimetypes are suitable at that point, you may just 
want to send a catalog of the URIs that indicate the languages 
within the entity and the client can decide what it can and can not 
process.

> I'm currently reading through the MT, specifically the RDFS entailment
> lemma.  It seems relevant.  Now, if only I had the "for dummies"
> version, I'd be all set. 8-/

I find stepping through the  entailment tests helpful sometimes:

    http://www.w3.org/TR/2003/WD-rdf-testcases-20030905/#tc_cert

Bill de hÓra

Received on Monday, 15 September 2003 05:18:52 UTC