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

Re: SEM DESIDERATA: my initial desiderata list

From: Pat Hayes <phayes@ai.uwf.edu>
Date: Thu, 4 Apr 2002 22:03:11 -0600
Message-Id: <p0510151db8d24c009dde@[]>
To: "Lynn Andrea Stein" <lynn.stein@olin.edu>
Cc: www-webont-wg@w3.org, "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
Sorry Ive been out of the loop for a while. Belated response:

>Several questions/clarifications:
>In http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0173.html, "Peter
>F. Patel-Schneider" wrote:
>>                  Potential Desiderata for the Web Ontology Language
>I (LAS) include only those potential desiderata on which I have something to
>>  Syntax:
>>  1/ Syntax is (equivalent to) n-triples (i.e., RDF syntax).
>>     All syntax (except, maybe, datatypes) is carried in triples.
>Pretty clearly we will need *some* sort of (unordered) listing/grouping
>construct that is bounded, i.e., defines a list/group containing *exactly* the
>listed/grouped elements.  daml:collection was one attempt to provide such a
>thing.  I don't believe we can avoid having such a thing.  (BTW, I also
>believe that adding such a thing to RDF's semantics may cause problems as it
>essentially introduces negation.  Pat?)

Heres one way to do it in RDF: have a special 'last thing' (say 
owl:LastThing) which when put into a container (which in RDF is 
essentially a sequence) indicates the 'end', like LISP NIL. However, 
unlike LISP NIL, it is only a marker; there could in fact be things 
'in' the container past this point - which is why it doesnt introduce 
negation -  ie the following would be consistent RDF:

ex:seq :_1 foo .
ex:seq :_2 owl:lastThing .
ex:seq :_3 baz .

In RDF this is an open-ended sequence of at least three things. In 
OWL it is a sequence of length one with one thing and a terminator. 
(Is baz in the container? It is as far as RDF is concerned, but not 
as far as OWL is concerned. The OWL sequence is an initial 
subsequence of the RDF sequence of the same name, as a general rule.) 
This would enable OWL to mark its ends and still encode its syntax in 
RDF triples.

>>  Semantics:
>>  1/ There is a model-theoretic semantics, compatible with the RDF(S)
>>  model-theoretic semantics.
>>     All triples are assertions.
>As we discussed, this is one several of us want to question.  But this is only
>a potential desiderata list....
>>     All URIs denote elements of the domain of discourse.
>This doesn't make them real, of course, but yes, they should denote
>"discussable" things; they are names after all.  (But we have to be careful
>about what it means to talk about unicorns....  forall x. (x a unicorn) -> (!
>exists x)......(or, in pseudo-n3,
>this log:forAll :x. {:x a :unicorn} log:implies {log:not {log:exists :x}}.

No, let us not go there. This is a rat-hole we should be careful to 
avoid. If we allow unicorns in the domain of discourse, then it is 
correct to say that unicorns exist. Existence just means being the 
referent of a name, and if unicorn names name unicorns, then unicorns 
exist. If you don't want unicorns to exist, then don't talk about 
The alternative that you are sketching is to allow names that refer 
in a sense (since they are names and the universals quantifiers range 
over their denotations) but also don't refer in a sense (since they 
arent 'real' and the existential quantifiers DONT range over them.) 
That is 'free logic', the logic you get when you allow nondenoting 
names, and it is a logical nightmare. All proofs get to be extra-long 
and are festooned with side-branches proving that the things that 
exist really do exist, which have to be there in order for the poof 
to be valid, since *any* name might be one of these unicorn names. Eg 
it would not be correct now to write 'pats phone number is 
8502024416', because "pat" might be a nonreferring name; you have to 
say "pat exists and his phone number is...." . You can't infer (P A) 
from (forall (?x)(P ?x)), unless you can prove that A exists, and so 

I would suggest that a much better way to say what you intend to say 
here is to have a predicate (class, if you like) of 'real things', 
and to simply say that unicorns are not in that class. But don't say 
that by negating the existential quantifier, PLEASE.

>>     All classes are elements of the domain of discourse.
>I'm not sure what this means.  Does this mean all classes mentioned?  All
>classes conceivable?  All classes describable (in some language)?  The class
>of all classes that do not include themselves as members?  I'm not trying to
>be difficult, just clear that we're on the same page.

I agree, this badly needs to be clarified. We certainly don't want to 
say that all conceivable classes are, and if we want a mechanizable 
logic we don't want to say that all describable classes are.


IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
Received on Thursday, 4 April 2002 23:03:10 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:56:42 UTC