URIs for terms: motivation [was: Requirements Document]

On Fri, 2002-02-08 at 11:37, Peter F. Patel-Schneider wrote:
> From: Dan Connolly <connolly@w3.org>
> Subject: Re: Requirements Document
> Date: 08 Feb 2002 08:55:47 -0600
> 
> > On Thu, 2002-02-07 at 13:45, Peter F. Patel-Schneider wrote:
[...]
> > > The document anticipates some technical features of OWL.  In particular, it
> > > uses URI as the term identification mechanism.
> > 
> > Again, to me, that's a straightfoward elaboration
> > of the requirement that we agreed to:
> > 
> > " A unambiguous term referencing using URIs"
> > 	-- http://lists.w3.org/Archives/Public/www-webont-wg/2002Jan/0089.html
> 
> True.
> 
> I am now officially unagreeing to the above.

OK, then I'll elaborate on why I proposed that requirement,
and explicitly clarified that the word "URI" should occur there.

Hmm... actually, this writeup explains it pretty well:

@prefix ed: <http://www.w3.org/2000/08/eb58#>.
@prefix dc: <http://purl.org/dc/elements/1.1/>.
"""
The fundamentals: The Universal Web

The most fundamental specification of Web architecture, while one of the
simpler, is that of the Universal Resource Identifier, or URI. The
principle that anything, absolutely anything, "on the Web" should
identified distinctly by an otherwise opaque string of characters (A URI
and possibly a fragment identifier) is core to the universality.

Great multiplicative power of reuse derives from the facts that all
languages use URIs as identifiers: This allows things written in one
language to refer to things defined in another language. The use of URIs
allows a language leverage the many forms of persistence, identity and
various forms of equivalence. Each language simply refers to the URI
spec - this is a flexibility point allowing the properties of naming and
addressing schemes to be defined separately.

"""

is ed:excerpt of [
  = <http://www.w3.org/DesignIssues/Architecture.html>;
  dc:title "Web Architecture from 50,000 feet";
  dc:date "Fri, 11 Jan 2002 02:29:39 GMT" ].

Jeff et. al., please consider citing/excerpting that stuff as
additional motivation for
"Unambiguous term referencing with URIs".


That text doesn't have any official standing, but (a) it's
easier to copy-and-paste than to re-explain over and over,
and (b) it is proposed to the TAG and under review.
cf. minutes 7Jan
http://lists.w3.org/Archives/Public/www-tag/2002Jan/0056.html


> > >  This brings up the URI vs
> > > QName discussion.  
> > 
> > Does it? Perhaps you mean to bring up the URI vs QName discussion,
> > but I don't see how the requirements document does.
> 
> True again.
> 
> I am now officially bringing up the URI vs QName discussion here.


FYI, the TAG has accepted a related issue:
http://www.w3.org/2001/tag/ilist#rdfmsQnameUriMapping-6

> My view is that WebOnt should be trying to heal the gap between XML and
> RDF, and coming down on the RDF side of the URI vs QName division is
> premature.

My view is that the use of QNames as anything other than syntactic
shorthand for URIs is a bug. I wasn't able to convince the XML Schema
WG, though there was considerable support for this view; i.e. that
there shoulnd't be multiple things that can have the same name
in the same schema.

The XML Schema WG acknowledged that
" We don't believe we have a good long term solution."
  -- http://www.w3.org/2000/05/12-xmlschema-lcissues.html#ids-not-names

Their rationale was:

"We declined to adopt the final part of your request, per (3) above,
because the impact on usability in the context of a design with six
nameable but clearly distinct kinds of components (type definitions,
element declarations, attribute declarations, attribute group
definitions, model group definitions, notation declarations) of
forbidding name collisions between these different kinds of things
would be too great.  In particular, it would have directly
contradicted the requirement that we reconstruct the facilities of
DTDs insofar as possible."
	-- WG response on LC121
http://lists.w3.org/Archives/Public/www-xml-schema-comments/2000OctDec/0058.html

Meanwhile, the importance of compatibility with DTDs is going
away. cf Clark's comments at XML 2001 Conference in Orlando:

@prefix ed: <http://www.w3.org/2000/08/eb58#>.
@prefix dc: <http://purl.org/dc/elements/1.1/>.
"""
His vision of a much-improved XML 2.0 included adding in XML Namespaces,
XML Base, and the Infoset, while subtracting DTDs and dealing with the
problem of character entities.
"""

is ed:excerpt of [
  = <http://www.xml.com/pub/a/2001/12/19/jjc.html>;
  dc:title "XML.com: Clark Challenges the XML Community [Dec. 19,
2001]";
  dc:date "Thu, 01 Jan 1970 00:00:00 GMT" ].


and the recent sketch of such an XML spec:

  * XML-SW, a thought experiment Tim Bray (Wed, Feb 06 2002)
  http://lists.w3.org/Archives/Public/www-tag/2002Feb/0031.html


So the use of QNames that don't directly correspond to URIs
is (a) not consistent with the principles that allowed
the web to grow to the point where it is, and (b) motivated
only by compatibility with technologies that aren't really
relevant to our task.

So I stand by my argument that "unambiguous term referencing using URIs"
is a requirement for our language design. If there
isn't consensus in the WG around this requirement,
then I don't mind exposing the issue to a wider
audience; i.e. (a) elaborating on why it's
a proposed requirement, as above (b) including the reasons
against making it a requirement, and (c) demoting
it from the list of 1st-class requirements to the list
of 2nd-class requirements, i.e. goals.


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Friday, 8 February 2002 15:28:35 UTC