Re: Determining what a URI identifies

Mark, on: Saturday, November 02, 2002 9:15 PM:

 > Section 2.5 of the arch doc, point 1 says;
>
> "The authority over a URI determines which resource it identifies."
>
> I believe that what a URI identifies is determined principally by *use*.
> The publisher is, of course, the authority on what it *should* identify,
> but it's up to them to ensure that they use the tools at their disposal
> to clearly communicate that to the world, lest the world think it
> identifies something else.  But in practice, I'd say, the world always
> has the final say, though with the proviso that the publisher has the
> power to change, given sufficient time, what the world believes it
> identifies.


I disagree.  It is the model for natural language
but not for specs. W3C and IETF (etc) specs determine what identifiers
identify and langauges mean.  if you allow arguments that misuse
changes the meaning then you open the whole stack to destruction
(by for example those who falsify email headers.)

> For example, I recall TimBL suggesting that
> "http://www.w3.org/Consortium" identified the W3C organization,

You remeber wrong.

http://www.w3.org/Consortium  identifies a web page.
The web page is about the consortium.

> and
> indeed a handful[1] of people use it this way.  But the vast majority[2]
> use "http://www.w3.org" for that purpose, and no information returned
> from a GET on the latter suggests that it isn't, so that isn't likely to
> change unless the W3C takes action.

Firstly you are confusing the identification of a document and the
identfication of the
subject of the document.  Secondly, you seem to thing that there should only
one
"real" identifier for W3C or for a page about W3C., neither is which are
true.

The meaning of a URI is determined by the URI spec, as Larry says.
That spec says you look in the IANA registry to find the spec for the
scheme.
The HTTP spec gives the authority through DNS to the domain owner.
So the ownership model is true for HTTP web pages
but only because the URI spec points indirectly to the HTTP spec.

 Larry, you gets this right, IMHO, in your mail  of 2002-1104T01:04, though
this
idea of me misusing the identifier above confused it a bit.

But then you go on in your next mail of Monday, November 04, 2002 10:26 AM
to say:

> RDF is a different story (as you know from the private
> conversation we've been having): RDF itself contains
> no grounding, and an RDF assertion that uses a URI
> could mean the actual resource identified, the thing
> described by it, or something completely different.

That is nonsense.  When RDF uses a URI it must
use it to identify the same thing
as in every other language.   RDF does have forms
of grounding of meaning - the first being the
sharing of the URI space with the rest of the web
languages, and the second being the definition
of terms in RDF schemas in RDF or natural language.
To refer to abstract objects, RDF is forced to use
a local identifier within an RDF resource
(real or imagines). This is explained pretty well in the
current RDF "Concepts" document [6]

> RDF
> doesn't say, it's the application using RDF that needs
> to specify how it is grounded. (In some ways this is
> as with syntactic elements in XML that are 'anyURI';
> whether the anyURI string is being used as a resource
> identifier or an indicator of something that has a
> different conceptual relationship to that resource
> depends on the application.)

The nice thing about RDF is that it *does* say.
Unlike XML applications which as you say use a
anyURI but don't tell you what relationship to that
URI is.  In a simple way, that is why RDF is good,
and why we call it the semantic web.
You know what is identified and the relationships
are explicit.

So when RDF talks about  <#joe> as having
a contact:mailbox of  <mailto:joe@example.com>
an RDF processor which is aware of the URI spec
and the spec of mailto:  knows that the object
is an email mailbox according to the email specs.

(((This of course drives the RDF logicians crazy as there
is no model theory of the email specs.  More generally,
logicians are used to defining a self-consistent
system in a vacuum and life in the big city^H^H^H^Hweb
can be relatively frightening.)))

> I'm uneasy about this part of RDF design, but I can
> accept it.

I think you misunderstood it and I hope you are
much happier with the actual case!


Tim

[6] Resource Description Framework (RDF):
Concepts and Abstract Data Model
W3C Working Draft 29 August 2002
a version of
http://www.w3.org/TR/rdf-concepts/

Received on Monday, 4 November 2002 14:17:20 UTC