W3C home > Mailing lists > Public > www-rdf-logic@w3.org > April 2002

Re: DAML Level of Effort for FY03-FY05

From: Drew McDermott <drew.mcdermott@yale.edu>
Date: Tue, 23 Apr 2002 17:32:30 -0400 (EDT)
Message-Id: <200204232132.g3NLWUq13970@pantheon-po04.its.yale.edu>
To: www-rdf-logic@w3.org

   [Didier Villevalois]
   > Shouldn't RDF support something like daml:imports in its core ??
   > daml:imports would be a subPropertyOf 'rdf:imports'.
   > As rdf enables to speak about the same things in different places, it needs
   > a such mechanism.

   [Dan Brickley]
   One problem with pushing this down into RDF's core is that the meaning
   isn't clear, and (worse) that the clarification of the meaning isn't
   something we can easily do with any of the formalisms currently being used
   for RDF, DAML+OIL, WebOnt.

   ...
   I don't find much use for XML-level inclusion mechanisms in RDF content,
   I guess because the imperative sense of "hey, you must go grab that XML
   and include it here" seems at odds with the largely passive, descriptive,
   propositional nature of RDF/DAML+OIL content.

Yes, I agree that sounds odd, but possibly for a different reason (see
below). 

   Another version at a different level of abstraction would be that the
   current document "in-some-sense-buys-into" the worldview, claims
   etc (including associated axioms etc) of the referenced
   document. This is hard to get right.

Au contraire.  It's extremely easy to get right.  

   I don't see any particular reason why such mechanisms need to be defined
   solely for ontologies/schemas, either.

No, but that's where we came in.  It's currently defined for
ontologies, and needs merely to be extended to datasets (sets of
descriptions using one or more ontologies).

   This is an interesting problem though. But until we see a clear definition
   of 'imports' that tells me what I'm doing when I say my doc imports
   another, I'm not interested in deploying that property in my data. Too
   risky! 

It's risky only if you use it recklessly.  For a nonreckless use,
suppose I am interested in checking a dataset for *inconsistency.*
(That's in fact the first thing we do with it.)  For this application,
getting as much information as possible from the component ontology is
exactly right.

   I'd also want to be able to import a frozen representation of a
   namespace or ontology (eg. with sha1/md5 or XML signature).

Sure, but that's an orthogonal issue.

                                             -- Drew McDermott
Received on Tuesday, 23 April 2002 17:32:35 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:42 GMT