Re: MISC: Internet Media Type registration: proposed TAG finding

From: Dan Brickley <danbri@w3.org>
Subject: Re: MISC: Internet Media Type registration: proposed TAG finding
Date: Fri, 24 May 2002 07:59:34 -0400 (EDT)

> so rdfs is...
> 
> > dangerously complicated.
> 
> ...because we said that the full meaning of a property can't be fully
> captured by current W3C-sanctioned formalisms?

No, because RDFSS includes langauge that is being taken to mean that RDFS
is supposed to incorporate the natural-language meaning of prose that is
the object of rdfs:comment statements.

> That's realism, not complication. There are many things that are true
> of some RDF properties that RDFS, DAML+OIL, even CycL couldn't readily
> capture.

Yes, precisely.  But that is *not* how the RDFSS specification is being
read.  It is being read as stating that these things *are* part of RDFS.

> If RDF property and class meaning is to be *exhaustively* captured by
> formalism, we'll need something more sophisticated than WebOnt's
> current sketch. If not, we need our specs to make clear that machines
> won't always understand the full picture.

Making it clear that machines do not always understand the full picture
would be much preferable.   Making it clear that RDF(S) is what the model
theory says that it is, and no more, would be even more preferable.

There is nothing wrong in shipping RDFS (or whatever) documents between
different applications and having those applications having side
agreements, perhaps expressed in rdfs:comment statements.  In fact, this is
unavoidable, at least for now.  The dream of the semantic web, that all
necessary meaning can be captured in semantic web formalisms, is just a
dream at this time.

However, either it is wrong to claim that the prose in these rdfs:comment
statements affects the RDFS meaning of the documents except insofar as the
literal string is the object of an RDF statement, or RDFS is a dangerously
complicated formalism.

> Dan

peter

Received on Friday, 24 May 2002 08:24:38 UTC