Re: Self-descriptive assertions

On Thu, Mar 25, 2004 at 12:56:09AM -0500, Bijan Parsia wrote:
> >I'm wondering if I had the wrong idea about what folks meant by
> >"assert".  I was assuming that in order to have the equivalent of a
> >text/calendar document, you'd need the graph plus a statement that the
> >graph was asserted.  But what I hear you and Dan saying is that the
> >equivalent is just the graph,
> 
> So I would assert.

8-)

I should have said before that this resolves my issue with the media
type.  Now I'm just trying to understand how all these concepts we've
mentioned fit together, if you don't mind helping me out with that.

> >This would mean that reification, parseType="literal", and using
> 
> Reification and literals prevent assertions from the point of view of 
> the graph

Right.

> >text/plain or application/xml, are all mechanisms that avoid making
> >mark:assertions simply by not yielding triples from some RDF/XML.
> 
> Where is it mandated that a consumer of application/rdf+xml parses it 
> to triples?

Fair enough.  What I meant to say was that from a message sender POV,
describing their RDF/XML document as text/plain is one way to avoid
communicating the graph.  If any recipient does extract the graph, then
that's "sniffing", and "bad" per the TAG finding on authoritative
metadata[1].

> I don't see that that is mandated by anything. It's 
> application specific. Internet explorer or mozilla rendering using 
> their XML view can be doing a right thing. A Relax ng validator that 
> checks application/rdf+xml documents against the relax-ng schema will 
> never parse to triples but is perfectly ok.

But a consumer of application/rdf+xml must be able to extract the
graph, otherwise they're lying if they say they accept that media type.
If they only do schema validation, they should advertise that they accept
application/xml.  They are still free to validate any */*+xml that they
receive, but the important part is that they're not committing to being
something that they aren't.

I believe my position to be consistent with [1], although it goes beyond
the scope of that finding.

> I feel pretty darn Q.E.D.y at this moment. I see no evidence anywhere 
> in any draft of the registration that suggests anything like what I've 
> read you to suggest. And I find the counterexamples legion and 
> compelling. *Even if* anyone found it important to try to specify, 
> right now, formal technical mechanism for connectng, oh, web retrievals 
> with assertings of RDF documents (or their graphs), the media type just 
> is *such* the wrong place to do it.

As I mentioned above, I'm content that my issue is a non-issue, and I
thank everyone for helping me understand this.

 [1] http://www.w3.org/2001/tag/doc/mime-respect.html

Mark.
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca

Received on Thursday, 25 March 2004 10:56:25 UTC