Re: [pedantic-web] Re: The OWL Ontology URI

On Tue, 2010-05-11 at 22:50 -0500, Pat Hayes wrote:
> On May 11, 2010, at 1:38 PM, Jonathan Rees wrote:
[...]
> > Suppose we have several things
> >  t1. An RDF/XML document / syntactic form / representation
> >  t2. An N-triples document / syntactic form / representation
> >  t3. An information resource that has t1 and t2 as "representations"
> >  t4. A graph that is serialized / encoded by t1 and t2
> > And we have a URI U such that GET U retrieves t1 and/or t2, and such
> > that U refers to t3.
> 
> Then we have already violated http-range-14. Because t1 and t2 are  
> information resources themselves, and moreover they are the sources of  
> the (REST-)representations that are returned with the 200-level http  
> code; so, U must refer to t1 (if that is what it GETs) or to t2 (if  
> *that* is what it GETs).

I don't know where you get the impression that things work that way,
Pat. I'll unpack this slowly and somewhat formally and see if
it gets us anywhere.

Jonathan's premises look like this, to me (reading "and/or" as "and"
for the sake of simplicity):

 _:t1 a :RDFXMLDocument.
 _:t2 a :NTriplesDocument.
 _:t3 a w:InformationResource; w:representation _:t1, :t2.
 _:t4 a :Graph; is :parse of _:t1, _:t2.
 _:U is log:uri of _:t3.
 _:t3 w:representation _:t1, _:t2.

and I have no problem with the supposition that:

 :RDFXMLDocument rdfs:subClassOf w:InformationResource.
 :NTRiplesDocument rdfs:subClassOf w:InformationResource.

from which we can conclude:

 _:t1 a w:InformationResource.
 _:t2 a w:InformationResource.

but I have no idea how you jump to:

   _:U is log:uri of _:t1.
or
   _:U is log:uri of _:t2.

In this and your previous message, you seem to suppose that

 { ?I is log:uri of ?r.
   ?r w:representation ?content }
  => { ?I is log:uri of ?content }.

but I don't know why you suppose that. The introduction
of webarch (http://www.w3.org/TR/2004/REC-webarch-20041215/#intro )
goes to some trouble to say that a URI that refers to ?r
doesn't (necessarily) refer to ?content.

>  So, it cannot refer to t3, because t3 cannot  
> be identical to t1 or to t2.
> 
> > We know that:
> >  - t3 has some of the same properties as t1 and t2 (e.g. authorship)
> >  - t3 is not the same as t1 or t2 (differs in some properties)
> >  - there are many IRs that could be t3, i.e. t3 isn't uniquely
> > determined by the above (e.g. could vary according to what *other*
> > representations exist, or by temporal behavior)
> >
> > The question is whether t4 = t3 is consistent with what has been
> > decided about web architecture, and/or about "good practice" (which I
> > think is what you're getting at). I don't see anything in what you say
> > that would force yea or nay.
> 
> I don't think this is the primary question. The basic point is that,  
> regardless of the status of t3 and t4, t1 and t2 are *definitely*  
> information resources, and so are required to be the referents of any  
> URI which returns 'representations' of them with a 200 code.

Not so.

>  The story  
> ends there, regardless of the informational status of graphs. As long  
> as graphs are not documents, and documents are information resources  
> that emit representations attached to 200 http codes, we are stuck.
> 
> >
> > If it's inconsistent, that is interesting because it ought to tell us
> > something general about IRs, which are otherwise mysterious. (Graph is
> > just a stand-in for any number of other boundary cases, such as
> > journal article or referent-of-data:-URI or DOM tree or XML element or
> > XML infoset.)
> >
> > If it's consistent, then why didn't they just do the obvious thing in
> > SPARQL and say that the graph URI refers to a graph?
> 
> Because it is kind of obvious that to do that would require graphs to  
> have names, and nobody was willing to admit the obvious fact that  
> SPARQL fits the named-graph view so much better than it fits what the  
> RDF specs say.

Something like that. I'll try to find records of the discussion.

> > That's why I
> > assume the SPARQL authors determined somehow that graphs are not IRs
> > (although Dan's right, they don't come out and say so and it's not
> > forced).
> 
> IMO, the entire notion of 'information resource' is both (a) broken  
> (in philosophy-speak, incoherent) and more important (b) completely  
> unnecessary. One can state the entire content of http-range-14 without  
> mentioning 'information resource'. I have done so for well over a year  
> now and will continue to do so.

That seems plausible to me; I came to a similar conclusion in 
some of my work (though I can't find it just now).

How do you prefer to state it?


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
gpg D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Wednesday, 12 May 2010 15:58:34 UTC