[sc34wg3] Type versus Class: settled yet?

[Murray Altheim]

[I hope I'm not too muddled -- pre-coffee...]

Mary Nishikawa wrote:
 > Sorry for cross posting, but this terminology needs some discussion by
 > WG3, too.
 >
 > *Bernard Vatant
 >
 >> >>BTW the use of type vs class should be made
 >> >>consistent at least with SAM prose, but I assume "topic type" here means
 >> >>"topic class")
 >
 > Actually, I think we still need to come to grips with these. In
 > Barcelona, we decided to replace the term "class" with "type."
 >
 > So now we should be using the terms topic type, subtype, and supertype.
 > While working on this draft, I found I really wanted to really use
 > "class"not "type." I don't think they are interchangable in all cases.
 >
 > Class of things is the set of things. Type of things is not, it is the
 > category that describes the things in the set. Please comment and
 > correct me if my thinking of the terms is wrong.

As with any terminology, its definition and usage is determined by
a community of interest, not by any sense of technical, moral,
aesthetic or other "correctness". In my readings in the history of
logic (and in current theory) I find a variety of such terminologies.
I related in a previous message some of the debates between Peirce,
Whitehead, and others on the topic. There's no final solution -- you
make a choice, and provide your definitions.

Your ice cream example is to me typical of the way most people use
the word "type", and I agree that "type" and "class" are not
synonyms. Most calculi use the word "class" to describe classes of
things, which is different than their types (self-described), and
different than collections of things (which are groups of individuals,
not classifications).

[...]
 > Date: Thu, 26 Dec 2002 18:15:38 -0500
 > From: Dan Brickley <danbri@w3.org>
 > Subject: Re: [topicmapmail] Superclass-subclass indentation in the Omnigator
 >
 > *Dan Brinkley
 >  >>Hey, if you folks haven't picked names for these two relations yet,
 >  >>I hereby propose 'type' and 'subClassOf'; type relates an instance to
 >  >>a class it is a member of, while subClassOf is a relation between
 > classes.
 >  >>FWIW these are the terms we use in RDF/S and OWL.

I'm comfortable with the meaning behind Dan's definitions. This
is in keeping with the original XTM 1.0 PSI set, which uses
"superclass-subclass relationship" and "class-instance relation-
ship". The word "type" is used only from the perspective of the
instance itself, hence we could have named the XTM element
<instanceOf> as <type>, though in the end think that would have
possibly been more confusing than not.

I very seldom find "supertype" or "subtype" in any modern books
or discussions on logic, which is part of the reason I didn't like
changing the XTM 1.0 PSIs to use those terms. IOW, I didn't
think this would have been an improvement in terminology, as
either being more accurate to the general understanding, more
common in usage, more pedagogically simple, or demonstrably
necessary. Programmers understand the concept of classes. I
also much prefer to keep the semantics of the existing PSI set,
as there are already a lot of applications that use it.

My vote would definitely be "superclass-subclass relationship",
"superclass", and "subclass", though adding the "Of" (as in Dan's
suggestion) to the name of the relationship rather than simply
naming it provides an unambiguous indication of direction in the
association, so I'd be in favour of that too. But the symmetry
of having both role names be included in the relationship name
is also pretty simple to understand.

    current:

       association name:  "superclass-subclass relationship"
       role 1:            "superclass"
       role 2:            "subclass"

    possibly:

       association name:  "subclassOf"
       role 1:            "superclass"  (or "isSuperclass"?)
       role 2:            "subclass"    (or "isSubclass"?)

Is the latter better? Not really, IMO.

But one might note that it's very likely that each community of
interest using Topic Maps will create their own PSI set so as to
make clear the specific semantics of their usage.

John Sowa posted this message yesterday to the Conceptual Graphs
list, which I think pertinent to our discussion here. He points
out that there will be large numbers of ontologies, and they will
change continually as needed by their respective communities.

-------- Original Message --------
Subject: CG: Re: SUO: Enlightened Semantic Web ;-)
Date: Thu, 20 Mar 2003 23:33:48 -0800
From: "John F. Sowa" <sowa@bestweb.net>
Reply-To: cg@cs.uah.edu
To: Chris Menzel <cmenzel@tamu.edu>
CC: Jean-Luc Delatre <jld@club-internet.fr>,     SUO <standard-upper-ontology@IEEE.org>, "cg@cs.uah.edu" <cg@cs.uah.edu>
References: <3e79140c.13613.0@bestweb.net> <20030320200116.GQ30186@tamu.edu>

Chris,

My recommendation has always been to have an infinite
number of ontologies, organized in a lattice.

CM> Agreed we should not fall into the error of making ontologies
  > static and changeless, but it seems to me an equal and opposite
  > error to allow them to be overly fluid.

Each ontology in the lattice is perfectly static -- it never
changes.  But it is quite easy to move to a neighboring
ontology, if that makes a better fit with reality or whatever
else you want to fit it to.

CM> It seems to me that a large measure of the success of the
  > semantic web, enterprise integration, etc will depend on ontologies
  > being quite dependably stable over time.  Once an ontology reaches
  > a certain point of maturity, while it should indeed be possible
  > to change and improve it, in line Jean-Luc's conception, it should
  > not typically be *easy* to do so; there need to be clear and strict
  > guidelines for ontological change.

Since we don't have much experience with the way ontologies
will come to be used among large groups of people and systems,
it is hard to make predictions about what will or should happen.

However, we (the computer industry and their users) have had half
a century of experience with libraries of subroutines and the ways
they have been used and abused.  Each such library is a procedural
encoding of something very similar to what we have been calling
an ontology.  Following are some observations:

   1. They never stop evolving until they die, i.e., fall into
      disuse and are no longer "maintained".  IBM has a charming
      euphemism for that state:  "functionally stabilized".

   2. It is indeed true that programmers depend on subroutine
      libraries to be "quite dependably stable over time".
      The way such stability is obtained is to bunch the changes
      into "versions", which are released periodically.

   3. Each new version is intended to be "backwards compatible"
      with earlier versions, but in practice, they inevitably
      cause many programs to "break" because of unexpected
      dependencies.

   4. Every operating system that has reached some level of
      maturity supports the option of having multiple versions of
      subroutine libraries installed simultaneously so that the
      old programs can link to the old libraries while the new
      ones can take advantage of the new libraries.

   5. One popular OS, which has never yet reached maturity, has
      created "DLL Hell" -- the option of having only one version
      of any DLL (Dynamic Link Library) on the system at any given
      time.  The result is that installing a new application from
      the manufacturer of that OS will automatically update any
      DLLs that it may require -- thereby causing other applications
      that use the same DLL to crash.

   6. An excellent way of avoiding DLL hell on those OSes that
      are based on DLLs is to avoid installing any software made
      by the same company that manufactured the OS.  This solution
      is successful because software from other companies installs
      new DLLs, but it never overwrites DLLs that come with the OS.

General observation:  I expect the evolution of ontologies to
recapitulate our experiences with subroutine libraries.  There
will inevitably be libraries of them with version numbers that
are periodically updated.  Any attempt to avoid having multiple
versions of the "same" ontology on the same system at the same
time will undoubtedly create a new Hell, not unlike DLL Hell.

John

========================================================================
To post a message, send mail to cg@cs.uah.edu.
To unsubscribe, send mail to majordomo@cs.uah.edu with the command
'unsubscribe cg' in the message body.
See http://www.virtual-earth.de/CG/cg-list/ for the mailing list archive
See http://www.cs.uah.edu/~delugach/CG for the Conceptual Graph Home Page
For help or administrative assistance, mail to owner-cg@cs.uah.edu


Murray

......................................................................
Murray Altheim                  <http://kmi.open.ac.uk/people/murray/>
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK

     "In Las Vegas Mr Gates also demonstrated a prototype
      fridge magnet which can be programmed to receive traffic
      reports, sports results and advertisements from local
      restaurants using the same FM signal as the wristwatch."
                                  -- The Guardian, 10 Jan 2003.