W3C home > Mailing lists > Public > www-tag@w3.org > December 2003

Re: 2.3 URI Ambiguity

From: Tim Berners-Lee <timbl@w3.org>
Date: Tue, 2 Dec 2003 10:39:29 -0500
Message-Id: <B7CD03B2-24DD-11D8-BFEF-000A9580D8C0@w3.org>
Cc: www-tag@w3.org, "Williams, Stuart" <skw@hp.com>, Mark Baker <distobj@acm.org>
To: Walden Mathews <waldenm@optonline.net>

On Dec 1, 2003, at 14:15, Walden Mathews wrote:

> It seems to me that the Architecture of the World Wide Web
> has to make a decision as to whether there is such a thing as
> indirect identification at the architecture level, and if so formalize
> what that is.

The problem was just that when we said "only use URI to identify the 
one thing", a lot of people responded with a counter-example of the 
person "identified" by their mailbox.  So the counter-example was only 
in so as to say, "No, we don't mean that".   If we have to formalize 
everything that we don't mean, then we would end up  extending the 
architecture document indefinitely!

The only way i can think of of making it clearer is to use say NTriples.

Direct identification:

  	here the same URI is use twice to identify the same thing,
	which is a person called fred.

	card:i rdf:type  contact:Person.
	card:i foaf:aka  "fred".


	Here the URI card <http://www.w3.org/> is used to
	identify two distinct concepts, with resulting confusion.

	<http://www.w3.org>  a  :Consortium.
	<http://www.w3.org>  dc:created "1994".

	<http://www.w3.org/> a :WebPage.
       <http://www.w3.org>   dc:created "2003-12-01".

Indirect "identification":

	_:x    rdf:type  contact:Person.
        _:x    contact:mailbox   <mailto:foo@example.com>.
	<foo.html>   dc:creator  _:x.

	_:y    contact:givenName "Frederick".
        _:y    contact:mailbox   <mailto:foo@example.com>.
	_:y     foaf:aka  "fred".

	contact:mailBox rdf:type owl:inverseFunctionalProperty.

	in this example, the fact that contact:mailbox is an inverse function 
property (erstwhile unambiguous property),  one can infer that the
unlabeled concepts _:x and _y are the same thing.  This is equivalent of
the way databases are typically used with keys.

(I don't think that this level of detail is necessary in the 
architecture document - it would be useful maybe in a finding.)


> Otherwise, beyond saying "don't do that" about using URI to
> identify more than one resource, I don't think there is anything
> to say, and one shouldn't try.
> Sorry to keep chiming in...
> Walden
Received on Tuesday, 2 December 2003 10:39:32 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:40 UTC