RE: Action item re URNs

-----Original Message-----
From: ext Tim Bray [mailto:tbray@textuality.com]
Sent: 02 June, 2003 04:26
To: WWW-Tag
Subject: Action item re URNs



Sometime in the next few minutes http://www.tbray.org/tag/URNs.html 
should appear; my long-promised note on the appropriateness of various 
URI schemes for XML namespace names.
-- 
Cheers, Tim Bray
         (ongoing fragmented essay: http://www.tbray.org/ongoing/)



Tim,

I think it would be good to address the issue of denotation
of URIs as it relates to their use as XML Namespace Names and
what representations might be obtainable via such URIs.

I.e. in addition to a URI used as a namespace name not needing 
to resolve to any representation -- it also need not denote an
XML Namespace. No semantic conditions are imposed by the XML
Namespaces spec on what an XML Namespace Name "means". It is
strictly a piece of syntactic machinery to allow one to "fake"
having URIs as element and attribute names without actually
actually having to change the XML syntax.

One can take a URI that denotes, e.g. an image of a poodle, and 
use that URI as an XML Namespace Name, and be in full compliance 
with both the letter and intent of the XML Namespaces spec.

Much discussion about XML Namespace Names, RDDL, and related
topics seems to have the inherent (and unfounded) presumption that 
every URI used as an XML Namespace Name does in fact denote an XML 
Namespace, which is definitely not stipulated by the specs.

Thus, if one is to actually have e.g. RDDL documents or other
such representations retrievable via XML Namespace Names, then it
should be made clear that there are some non-trivial semantic 
constraints on the choice of XML Namespace Names that
need to be taken into consideration -- to avoid introducing
ambiguity of denotation into the Web and Semantic Web.

One would not want the same URI being used to denote both
an XML Namespace and an image of a poodle.

Thus, folks should be clearly advised against using any URI 
that denotes anything other than an XML Namespace as an XML 
Namespace Name (e.g. one cannot use a URI that denotes a
vocabulary, or model, or schema, etc.) if they ever expect to 
provide representations of that XML Namespace Name via that URI.

And also, folks interpreting XML documents should be warned
that there is no guaruntee whatsoever that any URIs used
as XML Namespace Names actually denote XML Namespaces (or
anything at all) and thus should proceed with great caution
when attempting to obtain representations or descriptions.

--

I would (and do) personally advise folks to disregard 
XML Namespaces entirely insofar as semantics is concerned 
and deal solely with complete URIs, and use SW solutions such
as URIQA [1] to obtain knowledge about specific resources
(vocabularies, ontologies, terms, models, elements, schemas, 
stylesheets, services, instances, processes, etc., etc.)
irregardless of what punctuation is used in the serialization
of that knowledge or what (semantically irrelevant) substring
intersections one URI might have with another.

URIs are opaque insofar as their semantics are concerned. The
real problem with XML Namespace Names is that the XML specs
do not map qualified names to full URIs. An approach similar
to that taken by RDF would solve a lot of problems, and
emphasize the semantic emptiness of XML Namespace Names. It
would have been far better for XML to simply have allowed
full URIs as element and attribute names, and treated namespace
prefixes as syntactic sugar for ENTITY expansions.

Rather than positing any semantics for XML Namespace Names,
the TAG should be instead specifying how qualified names
in XML instances map to full URIs and thus diminishing any
need to concern oneself with XML Namespace Names at all, 
except as they relate to the particular syntax of XML.

I assert that URIQA fully alleviates any need to force
XML Namespace Names to have any semantic significance 
whatsoever, and makes approaches such as RDDL
obsolete and unnecessary -- allowing us to use the full
richness of RDF and the SW to publish and obtain knowledge
about resources (Web related or otherwise) and their
interrelationships without reference to the internal syntactic 
machinery of a particular serialization.

If one wants to know what a given URI (e.g. an element or attribute
name) in a given XML instance means, one can simply ask, using URIQA. 
One need not concern oneself about how the URI is punctuated in order
to meet the constraints of a given serialization. And by basing
semantics on full URIs, this unifies the web and SW architecture
irregardless of serialization, allowing applications to focus on
resources as denoted by URIs, not on how those URIs are expressed
in a given serialization (however prevalent to the Web).

Cheers,

Patrick

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

[1] http://sw.nokia.com/URIQA.html

Received on Monday, 30 June 2003 05:29:24 UTC