Re: Web Ontology Requirements Document - comments sought


I have a possible additional use case for you:  content negotiation.

Some time ago, I did some work for an IETF specification using logical 
expressions to describe sets of features, in a fashion having similarities 
with Description Logics.  In particular, the feature matching (i.e. 
matching content origination capabilities with receiver capabilities) was, 
I think, a form of subsumption calculation.  See RFC 2533 [1] for more details.

Now we have CC/PP using RDF to describe client capabilities, but without 
any framework for matching them with origination capabilities.  E.g. the 
sort of missing capability one could imagine is a number of content 
generation stylesheets, each associated with a description of required 
media features, from which a selection could be made by matching with the 
client CC/PP profile.  Preferably, this stylesheet would have a form very 
similar to the CC/PP profile.  This could provide a basis for protocol- and 
application-independent content negotiation [2].

A long-term goal of mine is to work to converge the CC/PP and IETF RFC2533 
frameworks to the point that the semantics, if not the syntax, are readily 
exchangeable.  Some work on converging vocabularies is underway.  I had 
been looking to the DAML/OWL work to provide a basis for implementing 
feature matching of descriptions using logical expressions.  There may be 
some difficulties with the RFC 2533 form, as it uses negation and 
disjunction, and keeps the computation tractable by limiting the 
expressions in other ways.  (I don't fully grok the constraints on DL 
expressions that are used to ensure tractability.)


I have one other comment on the requirements document, concerning 
procedural attachments.

While I can understand that there will be new applications that will need 
some kind of extension to OWL to fulfill their requirements, I think it 
would be a great shame if the language cannot satisfy its original use-case 
goals without such extension (particularly given that string and arithmetic 
primitives are also noted in the objectives).

Also, the term "procedural" makes me nervous.  I strongly feel that such 
attachments, where used, should not be used to allow an OWL expression to 
have side-effects, as that would introduce all sorts of 
evaluation-scheduling constraints (e.g. would effectively prohibit caching 
of already-evaluated results).  Related to this, I think that where 
external evaluations are used, the OWL processor should be able to know if 
they depend upon volatile conditions -- i.e. can the result be evaluated 
once and then cached for use as-needed?  (Specifically, I would like to be 
able to perform OWL evaluations in the framework of a functional 
programming environment using lazy evaluation.)




At 11:28 PM 3/19/02 -0500, Jim Hendler wrote:
>RDFers - the Web Ontology Working Group has published our first working 
>draft - a use cases and requirements document for the Web Ontology 
>Language [1]. We are soliciting comments on the document at 
> and general discussion of these 
>issues on rdf-logic.   We'd very much like to hear from you
>  thanks
>  Jim Hendler and Guus Schreiber
>  WebOnt WG CoChairs
>p.s. apologies for the cross post - we thought both groups might be 
>interested.  Please direct responses as discussed above.
>Professor James Hendler                 
>Director, Semantic Web and Agent Technologies     301-405-2696
>Maryland Information and Network Dynamics Lab.    301-405-6707 (Fax)
>AV Williams Building, Univ of Maryland            College Park, MD 20742

Graham Klyne

Received on Wednesday, 20 March 2002 08:09:48 UTC