W3C home > Mailing lists > Public > www-webont-wg@w3.org > October 2002

Re: on media types for OWL (5.13)

From: Dan Connolly <connolly@w3.org>
Date: 30 Oct 2002 16:58:30 -0600
To: "Peter F. "Patel-Schneider <pfps@research.bell-labs.com>
Cc: jonathan@openhealth.org, www-webont-wg@w3.org
Message-Id: <1036018714.7448.147.camel@dirk>

On Wed, 2002-10-30 at 08:07, Peter F. Patel-Schneider wrote:
> From: "Jonathan Borden" <jonathan@openhealth.org>
> Subject: Re: on media types for OWL (5.13)
> Date: Wed, 23 Oct 2002 22:36:22 -0400
> 
> [...]
> 
> > I'd really like a better story about how such a media type says anything at
> > all relevent to OWL.
> > 
> > Since OWL starts with <rdf:RDF> I fail to see why application/xml isn't just
> > as good as application/rdf+xml ***BUT*** if you can construct a story on how
> > the 'meaning' of a document isn't just a function of the meaning of a root
> > element, but rather depends on contained namespaces, then may be willing to
> > listen to this story ... of course I will hold you to that at a later point,
> > perhaps outside this WG if you catch my drift.
> > 
> > Jonathan
> 
> 
> (By ``validly reading'' below, I mean being read and understood, i.e., that
> the agent knows the correct handling of a document.  I believe that this is
> what the media type is supposed to support.)
> 
> The question is whether we (i.e., the W3C Web Ontology Working Group) want
> any agent (that is any agent, not just an OWL-aware agent) that is validly
> reading a Large OWL document with the following contents
> 
>  <rdf:RDF xmlns:rdf="...the usual..."
>   <rdf:Description rdf:about="http://foo.ex/bar#john">
>     <rdf:type rdf:resource="http://foo.ex/bar#Student">
>     <rdf:type rdf:resource="#Employee">
>   </rdf:Description>
>  </rdf:RDF>
> 
> to answer positively if it is asked whether the above document entails
> 
>  <rdf:RDF xmlns:rdf="...the usual..."
>  	  xmlns:owl="...the usual...">
>   <rdf:Description rdf:about="http://foo.ex/bar#john">
>     <rdf:type>
>       <owl:intersectionOf rdf:parseType="collection">
>        <rdf:Description rdf:about="http://foo.ex/bar#Student">
>        <rdf:Description rdf:about="http://foo.ex/bar#Employee">
>       </owl:intersectionOf>
>     </rdf:type>
>   </rdf:Description>
>  </rdf:RDF>
> 
> If so, we need to prevent agents that are only XML-aware or only RDF-aware
> from believing that they are validly reading such documents.

I don't think it's a matter of being "only RDF-aware" or "validly
reading"; in the general (and usual) case, agents only partially
understand documents; they can eliminate interpretations from
being models, but they don't usually end up with just one model,
and they don't assume that they know all the axioms/specifications
of all the properties in the document.

It's a matter of what basis you justify
your conclusions on. There's more than one
entialment relationship in the world.

>  The only way
> I can see of doing this is to have at least one OWL media type.

Hmm... the above example makes an interesting point in
that there's no path to "follow your nose" from the
premise to the owl spec; you could follow your nose
from the conjectured conclusion back to the owl spec,
if you were asked to prove the latter from the former.

So even in that case, I'm satisfied with just labelling the premise
and the conclusion app/rdf.

I'm starting to think that we really need
a proof/justification/explanation language in
order to specify the details of how all this work.

Since we've agreed to postpone that whole pile
of issues [oops; what happened to my action item
to raise that issue?], I can sympathize with the argument
for an OWL media type. (I'm not convinced
by it, but I have only intutive counter-arguments).

To re-terate/summarize: I propose to revise
the OWL reference (and/or guide?) to advise
users as follows regarding media types for owl:

=========
The following media types are available for
labelling owl documents:

	application/xml (as defined in RFC 3023,
		IETF Proposed Standard),
	application/rdf+xml (as defined in the
		RDF concepts/model-theory/syntax spec,
		which include/reference a media type specification,
		currently in draft form at
		http://www.aaronsw.com/2002/rdf-mediatype.html)
	applicaiton/owl+xml (as defined in a
		new section of our reference doc,
		and copied/excerpted into an
		Internet-Draft for IETF review)

Use application/xml to publish a document without
asserting it, i.e. as an example, test case, illustration,
object of study, etc.

Use application/xml+rdf or application/xml+owl to
actually say/assert the contents of the document,
i.e. to accept responsibility for it as if you
had translated it to natural language and
said/asserted the result.

Note that the connection between application/xml+owl
to this specification goes directly from the IANA
registry to this spec. The connection between
application/rdf+xml and this specification relies
on conventions established in [RDF concepts],
and is less mature. There is no specified
connection between application/xml and this
specification; such a connection must be
established by private agreement between sender
and receiver if the sender chooses to
label owl as application/xml.
=========


> 
> Peter F. Patel-Schneider
> Bell Labs Research
> 
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Wednesday, 30 October 2002 17:58:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:53 GMT