RE: DAML+OIL (March 2001) released

On Thu, 29 Mar 2001, Peter Crowther wrote:

> > From: Dan Connolly [mailto:connolly@w3.org]
> [...]
> > I argued (with Tim) against splitting the domains; but
> > tools like OILed require the split in order to do
> > efficient reasoning, and while I eventually want to
> > get beyond that requirement, the group agreed it's
> > worth keeping for at least a little while longer;
> > but we agreed to note that it's a stop-gap solution:
> >
> >   [[[
> >   RESOLVED: We will release an updated language release
> >   incorporating the current proposal, acknowledge the outstanding
> >   issues and concerns, and solicit feedback from the larger
> >   community.
> >   ]]]
> [...]
>
> It's worth thinking about what happens if the split is removed, though.  In
> particular, how does one avoid being able to express:
>
> - 5 is a kind of -3;
>
> - concrete class X is a kind of datatype D;
>
> - The datatypes ["A", "N") and {"Aardvark", 1, 3.14159} are equal;
>
> - refinements of data values ("the 3 which is the number of legs on a
> 3-legged pig"*) and what happens if you try to use the data values' built-in
> predicates, such as less-than and greater-than.  For example, is "the 3
> which is the number of legs on a 3-legged pig" greater-than "the 1 which is
> the number of legs on half a pair of trousers"*?
>
> Or does one simply define such concepts as incoherent?


However we define the language, we know for sure it will be a language in
which people can say stupid things, and in which they can design
ontologies that (how can I put this...) lack elegance.

The problem as I understand it is that the current proposal rules out the
definition of properties that have weak range constraints. For example,
the current Dublin Core practice of using 'http://purl.org/dc/elements/1.1/creator'
to point either to a typed resource or to literal data (ie. the creator's
name). The March release of DAML+OIL takes a strong dislike to this
practice (as do various Description Logic implementations, I'm told).

Allowing some properties to point to either resources or literals may be
bad modelling, but one might make the case that it avoids needless
duplication of property definitions for simple applications...

Dan

Received on Thursday, 29 March 2001 10:35:10 UTC