RE: ISSUE: owl:Class name misleading; try owl:Set?

I fully second the design with owl:Set and owl:sameMembersAs
and switched to it in http://www.agfa.com/w3c/euler/owl-rules
and can run the same 116 testcases as before, but easier

-- ,
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/


                                                                                                                       
                    Dan Connolly                                                                                       
                    <connolly@w3.org>        To:     Jim Hendler <hendler@cs.umd.edu>                                  
                    Sent by:                 cc:     Ziv Hellman <ziv@unicorn.com>, Jeff Heflin                        
                    www-webont-wg-requ        <heflin@cse.lehigh.edu>, "Peter F. "Patel-Schneider                      
                    est@w3.org                <pfps@research.bell-labs.com>, www-webont-wg@w3.org                      
                                             Subject:     RE: ISSUE: owl:Class name misleading; try owl:Set?           
                                                                                                                       
                    2003-01-03 07:09                                                                                   
                    AM                                                                                                 
                                                                                                                       
                                                                                                                       





On Thu, 2003-01-02 at 20:14, Jim Hendler wrote:
> At 18:44 -0600 1/2/03, Dan Connolly wrote:
> >On Thu, 2003-01-02 at 17:12, Jim Hendler wrote:
> >>  CHAIR'S NOTE
> >[...]
> >>  3. We made a decision at that owl:class is a subclass of rdfs:class
> >>  but not identical to rdfs:class (per closing of issue 5.22 and the
> >>  "semantic consensus"). The RDF Schema document [1] reads:
> >>
> >>  >RDF distinguishes between a class and the set of its instances.
> >>  >Associated with each class is a set, called the class extension of
> >>  >the class, which is the set of the instances of the class. Two
> >>  >classes may have the same set of instances but be different classes.
> >>  >For example, the tax office may define the class of people living at
> >>  >the same address as the editor of this document. The Post Office may
> >>  >define the class of people whose address has the same zip code as
> >>  >the address of the author. It is possible for these classes to have
> >>  >exactly the same instances, yet to have different properties. Only
> >>  >one of the classes has the property that it was defined by the tax
> >>  >office, and only the other has the property that it was defined by
> >>  >the Post Office.
> >>
> >>  so if we decided to go with a purely set based approach, the we would
> >>  not be a subset,
> >
> >No? Please explain.
> >
> >As far as I can tell, it's quite coherent to say that
> >the class of sets is a subclass of the class of classes.
> >That's what cyc does
> >http://www.cyc.com/cyc-2-1/vocab/math-vocab.html#Set-Mathematical
>
> umm, cyc says the sets of sets which contain themselves (collections)
> is a superset of the set of sets,

Not all cyc collections contain themselves. Most of the
useful ones do not, in fact.

> which isn't the same thing

It is the same thing.

A cyc collection completely analagous to an rdfs:Class.

And cyc says
           (genls Set-Mathematical Collection)

genls is analagous to rdfs:subClassOf. So this is tantamount
to

           Set-Mathematical rdfs:subClassOf rdfs:Class.

The explanation in
  http://www.cyc.com/cyc-2-1/vocab/math-vocab.html#SetOrCollection
says it pretty clearly...

"The second major difference between sets and collections is that no two
distinct sets can be co-extensional (be such that every element of one
is an element of the other and conversely). Sets are, so to speak,
identified on the basis of their extensions. Collections, on the other
hand, are identified by their intensional criteria for membership. So
collections which have exactly the same elements may be distinct,
differing in their respective membership criteria."

> >and KIF seems to do something similar... hmm...
> >taking another look, maybe this isn't relevant.
>
>http://meta2.stanford.edu/kif/Hypertext/node24.html#SECTION00073000000000000000

> >
> >But still... please explain why it's not OK to say
> >that owl:Set rdfs:subClassOf rdfs:Class?
>
>
> hmm, maybe this is where the misunderstanding lies, and perhaps
> someone else can say which of us has it right.  Kif and Cyc are
> distinguishing between Set and collections  (a mathematical entity
> which is a set allowed to contain itself) where we're talking about
> something different

we are?

>  - a class as an entity that can have properties
> associated with it other than just what the membership is.

??? I don't follow you at all. "can have other properties"
and "can contain itself" is descriptive of both rdfs:Class's
and cyc collections.


> As I read the semantics document, the class defined as "all people
> who are students and classmembers" is a different class than "all
> people who are students" even if we assert that
>   :classmember owl:sameClassAs owl:nothing.
> which would mean they necessarily have the same membership.  (because
> the restrictions and assertions associated with the class are
> different)
>
> Am I correct?

Not if they're owl classes; if they're owl classes, and they
have the same extension, then they're identical.

Hmm... double-checking current draft...
http://www-db.research.bell-labs.com/user/pfps/owl/semantics/rdfs.html#5.2

I don't see the constraint that IOT, IOC, and LV are
mutually disjoint. We talked about that in a telcon, no?
In fact, I asked that it be added to the owl schema:
           owl:Thing owl:disjointFrom owl:Class.
           owl:Thing owl:disjointFrom rdfs:Literal.
           owl:Class owl:disjointFrom rdfs:Literal.
and I thought Mike or Pat said they'd do it... or at
least think about it.

Hmm... I don't actually see the constraint that
if two classes have the same extension then they're
identical, but I also don't see anything that
conflicts with adding it.


>
> (This is fun, I'm being forced to read my way through the Semantics
> and work out issues - but perhaps those of you who understand this
> stuff better than I can say it better)
>
>   -JH
>
>
>
>
>
> >
> >>   but a different animal, and  we would have to reopen
> >>  this discussion and issue (which I'm not inclined to do).
> >
> >>  My belief at the moment is that Semantics is consistent with the RDF
> >>  S use of class, and thus I'm reluctant to introduce a different
> >>  semantics at this late date.
> >
> >I don't think I'm introducing a different semantics. I think
> >I'm just suggesting a more clear name for what we've got.
> >
> >>    I believe it would require a strong
> >>  (STRONG) consensus of the WG if the decision is made to change this
> >>  (esp. as the current also  seems to be consistent with DAML+OIL
> >>  according to Peter in [2]).
> >>
> >>  I may be misunderstanding the issue,
> >
> >One of us is.
> >
> >>   in which case I apologize, the
> >>  above are not meant to force a particular decision but to express
> >>  reluctance by the chair to introduce any major change on something
> >>  this basic.
> >>
> >>  -JH
> >>
> >>
> >>  [1] http://www.w3.org/2001/sw/RDFCore/Schema/200212bwm/
> >>  [2]
http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0011.html
> >--
> >Dan Connolly, W3C http://www.w3.org/People/Connolly/
--
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Friday, 3 January 2003 06:29:22 UTC