Re: on media types for OWL (5.13)

Dan Connolly wrote:

>
> On Tue, 2002-10-29 at 19:57, Jonathan Borden wrote:
> >
> > > >
> > > > I would say that application/owl+xml would indicate that, for
example,
> > for
> > > > Fast OWL, the OWL model theory were to be the defining model
theory --
> > this
> > > > could also be applied to Large OWL,
> > >
> > > No, you get that just from app/rdf.
> >
> > If that is the case, then app/rdf is fine. I don't entirely understand
how
> > this works, but perhaps this is a consequence of layering  for Fast OWL
as
> > well as Large OWL. When I asked for a 'better story' I am really asking
for
> > the logic that gets me to the OWL model theory from app/rdf.
>
> Well, I tried to give it.
> Maybe I'll find to give it again, with more clarity.
> But it's not quite a matter of "logic"; i.e. it's
> not really a 100% compelling rational argument; it's just
> a bunch of factors that I find persuasive. So I'm
> not very motivated to try much harder at it; either
> folks have similar intuitions or they don't.

I am pushing the issue because I have my own intuition that the answer is
important in and of its own right.

MIME media types are particularly lacking in guidance on how to derive the
semantics of, for example, a document with multiple namespaces. Sure we can
subclass app/xml into app/rdf+xml to direct that RDF/XML is a specialization
of the media type XML, but when we include multiple namespaces we
essentially need multiple inheritance. This is a relatively well known
limitation of media types among XML folk -- I know that there have been
threads on xml-dev if not www-tag regarding this issue. I have used the HTTP
extension framework to essentially expand on media type based content
negotiation e.g. http://www.rddl.org/httpext.

The reason that our 'solution' to this 'problem' is (IMO) important is that
it sets a precedent for assigning a semantics to a document. I was _hoping_
to be able to construct some story/logic that based the semantics of a
document as some function of its contained namespaces, but Peter's example
dusts that idea.

The example of an 'RDF' only document (at least as far as namespaces is
concerned) which has different entailments under RDF(S) and OWL is an
excellent one. Having an app/owl media type would allow the author to assert
(sic?) which entailments are licenced (of course the client may choose to do
what it pleases but such as how MIME media types are dealt with).

Unfortunately this doesn't answer the multiple namespace issue that I was
hoping to get an answer to, but if I can't motivate you to provide something
beyond intuition, and since I don't have an answer to this question myself,
particularly given Peter's example, I am swayed toward allowing either
application/rdf+xml or application/owl+xml.

Now a good argument _against_ application/xml for OWL or RDF documents that
aren't intended to be asserted is that the syntax/interpretation of the
fragment identifier is up to the media type and application/xml gives _no_
connection between #foo and rdf:ID="foo" (application/xml requires fragment
identifiers to be specified either with DTD attributes of type ID or XML
Schema attributes of type xml:ID -- RDF/XML has neither a DTD or XML Schema!
I don't support allowing application/xml _for that specific reason_.

Jonathan

Received on Wednesday, 30 October 2002 19:53:24 UTC