Re: RDDL Proposal from Tim Bray

I would like to ask two specific questions. These have been
asked before, in various ways, but I've yet to see a satisfactory
rationale/justification to the answers provided -- i.e. responses have
been assertions/opinions, not arguments with clear rationale or basis.

1. What is a namespace? What is its nature? What is its composition?

I'm presuming that the definition of a namespace is some interpretation
of the XML Namespaces rec, but many/most RDDL or RDDL-like proposals reflect 
what I would consider an invalid interpretation of that definition.

2. Is a RDDL instance a valid representation of a namespace?

This second question of course depends on the answer to the first. It
appears though, that a presumed (or desired) answer of "yes" to this
question is interfering with an objective answer to the first question. 
I.e., we will define a namespace to be whatever will allow a RDDL instance
to be a valid representation, even if that violates the REST model.

My own understanding/interpretation of the XML Namespaces rec is that
namespaces are simply collections of names; nothing more. And a namespace
does not include any knowledge about other resources which may describe
that collection of names, such as DTDs, XML schemas, RDF schemas, RELAX
NG schemas, etc. or style sheets to be used when displaying content
tagged with terms grounded in that namespace, etc.

Thus, a RDDL instance is mostly made up of knowledge that is not actually
part of the namespace resource -- and therefore has no business in any
representation of that resource. As such, I don't consider any of the RDDL
proposals (regardless of encoding used) to be valid representations of
a namespace resource and hence it is IMO invalid to return a RDDL instance
when dereferencing a namespace URI.

AS I've stated before, I fully agree that a standardized mechanism for
obtaining knowledge about resources relevant to other resources is
needed, but namespace URIs are the wrong mechanism to use to obtain
that knowledge, for numerous reasons that have been pointed out earlier.

C.f. http://lists.w3.org/Archives/Public/www-tag/2002Feb/0022.html

Resolving namespace URIs is a short term hack/kludge. There are better
ways to provide access to resource metadata, such as extending HTTP
to provide for e.g. META-GET which can return RDF statements about
a resource, without overloading the (limited) semantics of namespace
URIs in a manner that does not correspond to the relationships of
namespaces to terms to vocabularies to models to schemas, etc.

Regards,

Patrick

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


----- Original Message ----- 
From: "ext Tim Bray" <tbray@textuality.com>
To: "Tim Bray" <tbray@textuality.com>
Cc: "WWW-Tag" <www-tag@w3.org>
Sent: 03 December, 2002 03:13
Subject: RDDL Proposal from Tim Bray


> 
> > Here's the example.  There's a namespace whose name is 
> > "http://example.com/L".  It has a RelaxNG schema at 
> > "http://example.com/schemas/L.rng" with nature 
> > "http://relaxng.org/ns/structure/1.0" and purpose 
> > "http://www.rddl.org/purposes#validation".  It has a CSS stylesheet at 
> > "http://example.org/style/L.css" with nature 
> > "http://www.isi.edu/in-notes/iana/assignments/media-types/text/css" and 
> > purpose "http://www.rddl.org/purposes#render".
> > 
> > Please propose examples
> 
> I'm surprised at the lack of response to date.  The amount of 
> namespace-bearing XML in the world is increasing at a very high rate, 
> and that rate will accelerate dramatically sometime next year with the 
> release of MS Office 11.  I think it's important that we get some 
> consensus as to what those namespace canusefully point at, to provide 
> some interoperability in the marketplace.
> 
> So, here's my proposal.
> 
> Give up on XLink because this is not an end-user-oriented browsing 
> application, and because using "role" and "arcrole" for Nature and 
> Purpose is a kludge.
> 
> Give up on RDF because there is a poor match between RDDL's goals 
> (dereference a URI and use the results to look up other URIs in based on 
> nature & purpose) and RDF's goals (building an inference-capable network 
> of assertions about everything).
> 
> I propose one new element <rddl:rr> for each related resource, embedded 
> in XHTML as though it were in %Flow.mix, with attributes href, nature, 
> purpose, title, id, and prose.
> 
> <html xmlns:rddl="http://www.rddl.org">
> <head><title>The L namespace</title></head>
> <body><h1>The L Namespace</h1>
> <p>The name of the namespace is http://example.com/L.</p>
> <rddl:rr href="/schemas/L.rng"
>           title="RelaxNG Schema"
>           nature="http://relaxng.org/ns/structure/1.0"
>           purpose="http://www.rddl.org/purposes#validation" />
> <rddl:rr href="/style/L.css"
>           title="CSS Stylesheet"
>           nature=
>      "http://www.isi.edu/in-notes/iana/assignments/media-types/text/css"
>           purpose="http://www.rddl.org/purposes#render" />
> </body>
> </html>
> 
> Pro: easy for anyone to understand, parse, and generate.
> Con: Isn't XLink so if a generic XLink processor existed, it wouldn't be 
> able to use it.
> Con: Isn't RDF, so not directly part of the Semantic Web (on the other 
> hand, you could trivially generate any of the dozens of different 
> RDDL/RDF proposals based on reading the above).
> 
>   -Tim
> 
>

Received on Tuesday, 3 December 2002 02:55:48 UTC