Re: comments on http://www.w3.org/TR/2002/WD-rdf-schema-20021112/

>Brian & Pat
>I understand your examples. I think we are in agreement on 
>principles, bit still slightly divergent on terminology.

Indeed, we seem to be.

>
>From my perspective, the post office & tax office are obviously two 
>different contexts -- two different ways of viewing people.  I think 
>in terms of dividing all people into disjoint groups, according to 
>selected properties of the people.  In the first context, the 
>property is defined by the post office; in the second context, the 
>property is defined by the tax office.  The post office - tax office 
>distinction is somewhat clouded by the fact that both properties are 
>measures of physical location.

RDF (RDFS/DAML/OWL) does not have any notion of context, so your 
distinctions here cannot be rendered in these languages. And I 
confess to not understanding what it is that you mean by 'context' 
here, so it is hard to translate your intended meaning into RDFS 
terminology.

>You'll notice that I said "disjoint groups".  In a given context, 
>overlapping groups constitute an "ambiguous" division.

I think that thinking of classes as 'divisions' is potentially 
misleading. Subclasses of a given class need not constitute divisions 
of it. That way of thinking is appropriate for a taxonomy or 
classification tree, like the Dewey Decimal system, but not for 
general-purpose class reasoning.

>  For example, recall my previous email where Jane Doe belongs to 
>class "man" in Aristotle's context, but belongs to class "woman" in 
>Ms. Feminist's context.

Like most set theories, RDFS requires you to get this clear. Things 
either are in a class or they are not; there is no room for ambiguity 
or prevarication on the question. If you insist on making assertions 
which are only true in a context, and not true in others, then you 
should  probably not use RDFS, since RDFS content, when published on 
the Web, might be used in a very different context than the one for 
which it was written.

>
>I do not say that a class "is" its members.  I say that a class 
>"denotes" its members

That is not the sense of 'denotes' use in the RDF MT document.

>, or that the "meaning" of a class is its members, with all their 
>properties.  But that's somewhat beside the point.  The crucial 
>point is that the meaning of a class depends on context. 

Not in RDFS.

>Context is prior knowledge -- viewed from a particular perspective, 
>possibly characterized as a list of propositions.
>
>Back to the bottom line: defining "Class" in the rdf-schema document.
>1. I still say two different classes with the same members imply two 
>different contexts. 

Well, OK; you can of course say whatever you want; but that has no 
meaning in RDFS. If you try to apply this rule consistently, you will 
be forced to hypothesize rather a large number of contexts (o(2|n) 
where n is the number of classes, in the worst case.)

>It's your decision how to describe it.
>2. I have one other question related to the definition of "Class".  
>Is "Person" a subClassOf "Class", and "Dick McCullough" a 
>member/individual of "Class"? 

No.

>Or is "Person" a member/individual of "Class", and "Dick McCullough" 
>unrelated to "Class"?

Right.

>In other words, what is the meaning of "Person"? 

It is a class, containing people.

>And is there a difference between "Resource" and "Class"?

rdfs:Resource is a particular class, the largest one possible, the 
class of everything, the universe. All classes are subclasses of it.
rdfs:Class is the class of classes. All classes are members of it.
They are not the same class because there are things in rdfs:Resource 
that are not classes, such as Dick McCullough.

>You might add some words to the document to answer this question.

The MT does say all this quite explicitly. In an interpretation I, IR 
is the universe, and IC , a subset of IR, is the set of classes.

Pat Hayes

>============
>Dick McCullough
><http://rhm.cdepot.net/>knowledge := man do identify od existent done
>knowledge haspart list of proposition
>
>----- Original Message -----
>From: <mailto:bwm@hplb.hpl.hp.com>Brian McBride
>To: <mailto:rhm@cdepot.net>Richard H. McCullough ; 
><mailto:www-rdf-comments@w3.org>www-rdf-comments@w3.org
>Sent: Wednesday, November 27, 2002 8:25 AM
>Subject: Re: comments on 
><http://www.w3.org/TR/2002/WD-rdf-schema-20021112/>http://www.w3.org/TR/2002/WD-rdf-schema-20021112/
>
>At 06:02 27/11/2002 -0800, Richard H. McCullough wrote:
>>Yes, I am assuming that two classes with the same members are the same class.
>>
>>If that is not true of rdfs:Class,
>
>As I wrote before, it is not.
>
>then either
>>1. you are talking about "currently known members" of a class
>
>I don't believe that to be the case, as RDF makes no closed world assumptions.
>
>>or
>>2. you are talking about two "different contexts",
>
>I don't believe that to be the case either, as we don't define a concept
>called 'context'.
>
>>i.e., two different ways of viewing the same individuals,
>
>Maybe.
>
>Consider the rdfs:Class A, hypothetically defined by the post office to be
>the rdfs:Class of people whose address has the same zipcode as mine, and
>the rdfs:Class B hypothetically defined by the tax office to be the
>rdfs:Class of people living at the same address as me.  As it happens,
>classes A and B have the same members.
>
>A and B have different properties; they are different things.  rdfs:Class A
>has the property that it is described in a document
><http://example.org/schema/postoffice>http://example.org/schema/postoffice, 
>a property that is not true of B.
>rdfs:Class B has the property that it is described in a document
><http://example.org/schema/taxoffice>http://example.org/schema/taxoffice, 
>a property that is not true of A.  A
>and B are different things.  They just happen to have the same members.
>
>Let me try and bridge the gap here.  There are two different concepts
>floating around.  I think you call them context and class where a class
>*is* the set of its members.  Two classes with the exactly the same
>membership are the identical.  RDFS has a similar but different model.
>
>In RDFS we have two concepts.  There are rdfs:Class's each of which has an
>associated set that is the set of its members.  But the rdfs:Class is not
>the same thing as the set of its members; to use a term I first heard from
>a colleague that I particularly liked, the set of its members is *nearby*.
>Thats just how it works in RDFS.
>
>We are not going to get very far if we get your concept of class mixed up
>with what an rdfs:Class is.
>
>Now I figure we might take this in steps:
>
>    1.  You need to understand what the RDFS model is, and in doing so, we
>need to understand from you what bits of the specs are not clear to you.  I
>think we got the message we should explain the notion of class better :)
>
>    2.  Having figured out what RDFS really is, then you should check it for
>flaws.
>
>You should know that we are (I hope) pretty far down the process of writing
>the spec.  We have decided all our issues and are now just writing them
>up.  What that means is that for comments of the form:
>
>    o Hey, I've got a better idea; its too late for this round - a future
>working group will need to look at those.  You have to understand, if we
>didn't do that we would never get finished.
>
>    o Hey, thats broken because ... we will deal with, though we have to
>hope there will not be too many of those.
>
>>with two different sets of relations to other classes.
>>
>>If 1. or 2. is not true, then you're not talking about reality.
>
>I am writing about RDFS.  Draw your own conclusions.
>
>Has this helped at all?
>
>Brian


-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam