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

Re: Proposal to close issue 5.21

From: Evan Wallace <ewallace@cme.nist.gov>
Date: Thu, 24 Oct 2002 12:03:41 -0400 (EDT)
Message-Id: <200210241603.MAA01795@clue.msid.cme.nist.gov>
To: www-webont-wg@w3.org

Peter Patel-Scheider wrote:

>To further this,
>	_:x owl:disjointUnionOf [_:d1 ... _:dn] .
>has the same meaning (ignoring the RDF triples that arise from the syntax) as
>	_:x owl:unionOf [_:d1 ... _:dn] .
>	_:di owl:disjointWith _:dj . 		1<=i<j<=n
>The only thing going for owl:disjointUnionOf is that it uses fewer triples
>than the alternative.  However almost all disjoint unions are small so the
>number of owl:disjointWith triples will not be that large.

This is pretty much what I asked for in Bristol.  Because of this, (and 
because I couldn't find anyone at NIST who objected to it) I expect
to concur with the recommended issue resolution.  However, I don't believe
that statements such as the last sentence above are very meaningful.  How 
do we know the set of "all disjoint unions"?  What is "small" in this
context?  What does "almost all" mean?

In an EXPRESS model for cutting tools [1] which is a precursor to an ISO 
standard for same, I found 34 occurances of the the equivalent EXPRESS
syntax for disjointUnionOf, ONEOF.  The average size of the class lists
which were the object of these statements was 4.  I would agree that is
small.  However, the largest size was 11.  That expands to quite a few 
more disjointWith statements than I would want to write by hand. Which 
leads to the following.  

What is clear to me is that some n-ary form of Disjoint will exist in 
the presentation syntax of tools that create OWL Ontologies.  Even in 
Dan's proposed example of this for our GUIDE document he left the 
expansion of his small list to the "editor/reader".  It would make sense 
to include such a construct in our presentation syntaxes.  It is already 
the natural interpretation of the predefined Disjoint constraint in UML 
which applies across a set of generalization relations (subtypeOf).

[1] http://www.mel.nist.gov/rrm/fy97/jul97mrmodel.exp


Evan K. Wallace
Manufacturing Systems Integration Division
Received on Thursday, 24 October 2002 12:03:51 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:56:48 UTC