Re: DAMl "Thing" should be Top, Universal class - including concrete types

In response to the question of versioning -
there are two aspects to consider:
1- versions of the knowledge representation system
2- versions of the ontologies built in the KR language.

On 1, we had complete control - we were the central source for all 3 versions
(lisp, c, and c++) of CLASSIC.  There were a number of universities who made
extensions but they either sent their extensions to us for incorporation of the
future releases, or they handled the updates.
We worked very hard to only make monotonic changes to the language so that
people could have their extensions work with new releases.

On 2, we had very little control.  Application developers make their changes as
they need to.  Most smart ones who are working on systems that are to be
maintained by people and expected to have long life cycles only make monotonic
changes, but as we all know, very few people get their ontology designs "right"
the first time and thus, in most applications, there were non-monotonic changes
to ontologies.
We did have some applications where there was distributed ownership of domain
ontologies.  Nothing like what we see already on the web  and what we will see
soon.
Thus, you could be making the point that one application may want a range of say
strings on a property (say booktitle) and another application may want a range
of
of a structured term (that say has a string component for actual title in
english, another string in french, etc.)
My response to this is that one can use a property like booktitle-string  and
another use a property like booktitle.  booktitle might have a range that is a
composite object that contains a property called booktitle.

Deborah


Peter Crowther wrote:

> > From: Deborah Mcguinness [mailto:dlm@ksl.stanford.edu]
> > I would posit that we have not yet seen a compelling
> > naturally occurring
> > example of a need for combination of the types.
>
> Deborah, how much control did you have over versioning within CLASSIC and
> the systems you built around it?  I agree with you for CMs where there is a
> centralised create-distribute-modify-redistribute approach to versioning.
> *The* problem with the Semantic Web, IMHO, is going to be the same as *the*
> problem with the HTML Web: no central authority can control the version of
> anything that is in use.  This problem applies to standards, and gives Web
> authors the world over nightmares of the
> should-I-support-pre-CSS[2]-browsers kind; but, worse, it applies to
> distributed models too.
>
> At its best, this may subject us to the logical equivalent of link-rot,
> where the targets of URIs are either inaccessible or deleted.  At its worst,
> it leads to part of a user base describing the world using one version of a
> model where (say) book titles are strings, and another part of the user base
> describing the world using another (later) version of the same model where
> (say) book titles are complex.  If both publish their portions of the model
> using RDF, how can we reason about both at once?  Or is there a bodge that
> treats the two versions as semi-independent?  Or can we not reason about
> both at all?
>
> I'm sure exactly the same problem has cropped up with XML, where the DTD
> referred to in the DOCTYPE has changed since the data was created and the
> data is therefore invalid.  However, I suspect the solution there ("it's
> still XML, so just parse it") is not one that we can apply here.
>
>                 - Peter
> --
> Peter Crowther

--
 Deborah L. McGuinness
 Knowledge Systems Laboratory
 Gates Computer Science Building, 2A Room 241
 Stanford University, Stanford, CA 94305-9020
 email: dlm@ksl.stanford.edu
 URL: http://ksl.stanford.edu/people/dlm/index.html
 (voice) 650 723 9770    (stanford fax) 650 725 5850   (computer fax)  801 705
0941

Received on Tuesday, 6 February 2001 17:17:44 UTC