RE: Valid representations, canonical representations, and what th e SW needs from the Web...

> -----Original Message-----
> From: ext Bullard, Claude L (Len) [mailto:clbullar@ingr.com]
> Sent: 31 January, 2003 17:09
> To: 'Sandro Hawke'
> Cc: www-tag@w3.org
> Subject: RE: Valid representations, canonical 
> representations, and what
> th e SW needs from the Web... 
> 
> 
> 
> Perhaps where one sees resource, if one substitutes 
> "virtual signified", it is easier to understand. 
> I don't think it changes one wit the way the 
> system works.  On the other hand, if the problem 
> here is conflicting models, it is difficult to 
> conceive of the semantic web and the traditional 
> web as the 'same' system; they appear to be 
> systems that share signficators and indirectly, 
> signifieds.
> 
> For that to work in the normative RFCs, the 
> URI RFC needs to be syntax only, then other 
> RFCs or specificationd deal with the affective 
> model.

Well, I think that one major hinderance has been the
use of http: URIs to denote non-web-resources. I see
there being an inextricable semantics embodied by all
http: URIs as denoting web resources just as all
mailto: URIs denote mailboxes.

There seems to be some kind of perverse aversion to
allowing URI schemes to impose semantics on their
denotation, which puzzles me to no end, as it seems like
a clean and effective way out of this mess.

Web resources are denoted by http: URIs and you get
representations of them. Other resources are denoted
by other kinds of URIs, and you can relate web resources
to them in order to get representations of them. In
the case of digital resources, those representations
can/should be canonical (bit-equal copies).

Thus:

   <voc://example.com/me> a foaf:Person .
   <voc://example.com/me> x:representation <http://example.com/me.html> .
   <http://example.com/me.html> a x:WebPage .
   <voc://example.com/me> x:representation <http://example.com/me.jpg> .
   <http://example.com/me.jpg> a x:Image .
   <voc://example.com/me> rdf:seeAlso <http://example.com/me.rdf> .
   <http://example.com/me.rdf> a x:KnowledgeBase .

etc.

HTTP and other protocols can provide representations of
resources denoted by other URI schemes than http: but
the semantics of the denotation of those schemes may
be constrained by the definition of the scheme. So if
you have a very clever server, which can return some
representation of <voc://example.com/me> by some 
heuristics of determining what optimal representation
to give, fine, but that's simply a kind of redirection.

The voc: URI still denotes a non-web resource, even though
one might get a representation of a web resource when
dereferencing it.

(in fact, if folks treated the return of RDDL instances
when dereferencing XML Namespace URIs as default redirection
to another resource rather than as representations of 
the namespace resource itself, I'd be alot more comfortable 
with it)

But as soon as you try to use http: URIs for everything, for
both web resources and non-web resources, things break down,
since you can't talk about two things having the same URI
and thus cannot talk both of concepts and web resources
providing representations of those concepts.

Cheers,

Patrick

--
Patrick Stickler, Nokia/Finland, (+358 40) 801 9690, patrick.stickler@nokia.com

Received on Friday, 31 January 2003 11:50:16 UTC