Enclosed are comments forwarded with permission from the protege group.
I attend some of the protege group meetings and received some verbal
comments after our introduction of OWL Lite and the Feature Synopsis
document and I asked if they might be able to collect their thoughts in
an email.
This is the email collected by Ray Fergerson.
==================
--
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
(voice) 650 723 9770 (stanford fax) 650 725 5850 (computer fax)
801 705 0941
Forwarded message 1
Deborah,
Just getting back to you with some comments on OWL and OWL Lite.
The summary is that we feel that there is a need for a semantic
web language like the OWL Lite subset of OWL with a feature set that
is familiar for those with a modeling background. We think it will
also be much easier for us and other tool suppliers to support the OWL
Lite subset with a simple, intuitive user interface. While our
goal will eventually be full OWL support, this will require much more
work and, realistically, will be much more complicated to understand and
work with. The OWL Lite subset has tremendous value by itself, both
for ourselves (as a tangible implementation step) and for our users.
I have included below a list of areas where we still have problems
supporting some things in OWL Lite. Several of these are trivial and
some are more substantial. Nevertheless we would like to make it clear
that our group was pleased generally with the OWL Lite spec and we
feel that it is an improvement on its many predecessors.
Ray (for the Protege group)
-------------------------------------------
Areas where we will still have trouble are:
(1) Protege has union semantics for both domain and range on slots. We
can't get this with OWL Lite. This is a problem. There are workarounds
but they are ugly.
(2) Protege doesn't have any easy way to express the notion that two
instances are equivalent. It is unclear how to add this capability.
(3) The cardinality restrictions in Owl Lite still strike us as strange.
This seems to be an area where the difficulties in reasoning about a
feature (as a result of (2) above) is preventing a desirable feature
from being included in the language. This feature (or lack of a
feature) is not a big problem for us but I suspect that it will
come up again and again as you try to popularize the language.
(4) We don't have any way to express "someValuesFrom". This has come up
before but, as far as I am aware, the only people who have ever requested
this are trying to map Protege to the feature set of another existing
system. No one has ever wanted it for actual modeling. (I really object
to implementing these "me too" features that make things complicated
but which no one really wants.)
(5) Protege doesn't support the properties "transitive" and "symmetric".
These wouldn't be too hard to add though and they are requested
occasionally by our users doing modeling.