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

Re: proposal for last session of July face-to-face (new issues?)

From: Dan Connolly <connolly@w3.org>
Date: 19 Jun 2002 12:49:22 -0500
To: "Peter F. "Patel-Schneider <pfps@research.bell-labs.com>
Cc: www-webont-wg@w3.org
Message-Id: <1024508963.26122.131.camel@dirk>

On Wed, 2002-06-19 at 10:03, Peter F. Patel-Schneider wrote: 
> From: Dan Connolly <connolly@w3.org>
> Subject: Re: proposal for last session of July face-to-face (new issues?)
> Date: 19 Jun 2002 09:29:02 -0500
> > On Wed, 2002-06-19 at 09:05, Peter F. Patel-Schneider wrote:
> > [...]
> > > These leave three portions of OWL unspecified:
> > > 
> > > M1 What is an OWL KB in triple form, and, more importantly, what
> > >    collections of triples are not OWL KBs?
> > 
> > I expect all collections of triples are OWL KBs; that's
> > the way I read the DAML+OIL Note. It looks like a
> > new issue belongs on our issues list.
> Well if all collections of triples are OWL KBs then there are some 
> problems.

Yes; there's no free lunch... 

> For example, what is the meaning of a malformed list, i.e., a list 
> 1/ where some list node does not have a rest,
> 2/ where some list node does not have a first,
> 3/ where some list node has two first,
> 4/ where some list node has two rests,

Those don't seem hard: the latter two are contradictions, 
and the former two just mean there aren't enough premises 
or whatever to come to interesting conclusions. 

> 5/ that is circular, or
> 6/ that is infinite.

I see two options: 

(a) do without lists altogether, ala writing 
(and P (and Q (and R S)))) 
in stead of (and P Q R S). 

That would be tedious; oneOf would go from a simple 
binary relation to an ugly thing ala onProperty/hasValue. 

(b) work thru the details of induction over lists. 
This would take us outside of FOL, which would 
make the formal aspects of our spec more complex, 
but most users would probably prefer it 
to the tedious option (a). 

Lots of practical systems do this: Larch 
and ACL2 call themselves 1st order systems, 
but they sneak in induction.

Well, ACL2 is much less sneaky...

  A Precise Description of the ACL2 Logic

But Larch is pretty glib...

"Except for the fact that it provides a rich set of notations for
functions, LP is based on the syntax and semantics for first-order logic
found in many textbooks."
  -- LP, the Larch Prover -- Logical syntax and semantics

(I still need to study Michael's stuff on ACL2 and OWL...
http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0301.html )

> For example, what is the meaning of a malformed restriction, i.e., a
> restriction 
> 1/ that is missing an onProperty element,
> 2/ that has none of the other parts of a restriction, or

again, it's just as if you have a rule that says 
IF: parent(X, Y) 
AND brother(Y, Z) 
THEN: uncle(X, Z) 
and you have brother(a, b) with no clues about 
who a is a parent of; you just can't come to any 
uncle(X, b) conclusions. 

> 3/ that has more than one of the other parts of a restriction.

> These are a problem for the DAML+OIL way of expressing the model-theoretic
> semantics of triple KBs, but there is a solution there because the DAML+OIL
> model theory keys off the (syntactic) triples, not the (semantic) IEXT.
> [...]
> > > M3 What is the formal meaning of an OWL KB?
> > 
> > Whatever we specify it to be, no?
> [...]
> > > 	http://www-db.research.bell-labs.com/user/pfps/owl/translation.html
> > 
> > There's also my proposal to do a model theory straight from
> > the triple abstract syntax:
> > 
> >   * layering (5.3,5.10): a same-syntax model theory
> >   Dan Connolly (Thu, May 30 2002)
> >   http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0264.html
> Your proposal is not a model theory from the triple syntax but is instead a
> model theory from RDFS interpretations.  Your way of specifying semantics
> for triple KBs is *much* harder than specifying semantics from the triple
> syntax.  In particular, it cannot handle malformed lists and restrictions
> in the way that they are handled in the DAML+OIL semantics. 

Yes, specifying exactly how lists work is tricky. We could perhaps avoid
that. But I think it's worth the effort, since the effort is over 
when we finish our spec, but the effort for users to deal with a klunky 
language lasts long after. 

> > Dan Connolly, W3C http://www.w3.org/People/Connolly/
> Peter F. Patel-Schneider
> Bell Labs Research
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Wednesday, 19 June 2002 13:49:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:56:45 UTC