RE: What to do about namespace derived URI refs... (long)

> -----Original Message-----
> From: ext Sean B. Palmer [mailto:sean@mysterylights.com]
> Sent: 07 June, 2001 00:47
> To: Seth Russell; Patrick.Stickler@nokia.com; www-rdf-interest@w3.org
> Cc: Ora.Lassila@nokia.com
> Subject: Re: What to do about namespace derived URI refs... (long)
> 
> 
> > [...] you have stepped over the line between being
> > practical and being pig headed.
> 
> Seth!
> 
> > When a person is at the point of naming an ideal or real
> > entity that cannot be  accessed on the Internet; they need
> > a simple way to coin the URI. The method they use should
> > guarantee that they will not be colliding with other Internet
> > behavior such as bookmarking.
> 
> Fair enough. But the URI reference that one creates for RDF are
> usually FragIDs within some RDF document, not an HTML document, so
> this should be practically fine. 

HUH?!

"usually" doesn't cut it.

The problem is that folks are using various MIME content types to
define the identity of components belonging to abstract ontologies, blurring

the distinction between the concept and the definition, and the URI
reference into those definition schemas is specific to the MIME
content type of the definition, and hence *not* portable. Thus,
when I use XML Schema to define my properties and their allowed
values -- and then try to use RDF Schema to define additional properties
and relations about that ontology, I have *no* consistent mechanism
of reference because the fragment syntax for XML Schema and RDF 
is not the same!

Then later, when some better schema encoding is chosen over XML
Schema, and I wish to use it instead for defining the serialization
of statements based on my abstract ontology, all of those nice
RDF statements using the XML Schema URI refs must all be changed
to use the fragment syntax of the *new* schema definition.

Yet, all along, I am just talking about abstract concepts in an
abstract ontology. The schema and serialization encoding is irrelevant
except for getting my knowledge into my knowledge base (triples)
or moving them to some other knowledge base. Why is it then
dominating my reference schemes?!

It is the tail wagging the dog.

> For exmaple:-
> 
>    http://robustai.net/seth.html#Truth
> 
> It's an HTML page, you choose to define "#Truth" as some bits of data
> in that page. That's absolutely fine, and consistent with how the Web
> works and what have you.
> 
>    http://robustat.net/set.rdf#Truth
> 
> That's an RDF document, with a FragID of "#Truth" after it. As such,
> you can't browse it conventionally as you would HTML, and it contains
> data, not documentation. By adding ID="Truth" your browser won't go to
> that FragID. Hence, you can use that to identify your concept of
> "Truth", and it won't conflict with any bookmarking programs, because
> no one is going to bookmark it - they can't even conventionally access
> it.

Yes, but these two do not reify "Truth"! Both URLs simply
constitute rdfs:seeAlso or rdfs:isDefinedBy (or some other)
relations!

You have not yet defined a consistent reification of the 
concept *Truth*! 

If you wish to use a URL ref as the identity of an abstract concept 
with an RDF Schema URL ref, as per http://robustat.net/set.rdf#Truth,
fair enough but *beware* that you then can never reify that
abstract concept using any other MIME content type without risking
the obsolescence of every statement using that URL ref as
the subject identity (and the fragment syntax may be different, or
have a different interpretation), nor can you make any statements whatsoever
about the schema *definition* reifying the abstract concept, as then
there would be ambiguity as to whether you mean the abstract 
concept or the concrete definition of it within an RDF schema.

I actually don't see how one can differentiate between the XML
interpretation of names in an RDF(XML) Schema instance versus 
the reference of rdf:IDs within that schema instance. Is there
any normative definition of this? Neither the RDF nor RDF Schema
specs specify that they have distinct MIME content types other
than text/xml, so on what do you base your argument that a browser
would interpret a fragment reference for RDF any differently than
XML, which uses XPointer, which interprets "#Truth" just as HTML
does now??? Are MIME content types for RDF and RDF Schema defined
elsewhere then the specs?

---


Abstract is abstract. Concrete is concrete. The soul of the
semantic web is the intersection of the abstract as it references
the concrete. All we have so far is the concrete, which by
smoke, mirrors, and just plain good luck is being passed off
as abstract in the various toys and demos here and there, because
URL refs can be hacked and coerced to work like URNs.

(I'm surely stepping on a lot of toes with that statement, but
maybe a few other folks would like to join the conversation and
demonstrate my error and ignorance ;-)

Eh?

Patrick

Received on Thursday, 7 June 2001 04:46:32 UTC