W3C home > Mailing lists > Public > www-webont-wg@w3.org > September 2002

Unique Names Assumption & consequences for implementation / Guide document

From: Raphael Volz <volz@aifb.uni-karlsruhe.de>
Date: Mon, 16 Sep 2002 10:20:05 +0200
To: "Webont" <www-webont-wg@w3.org>
Message-ID: <DMECLAFLIOFJEFFIAJPCGENPCPAA.volz@aifb.uni-karlsruhe.de>

Hi -

I'ld like to see some discussion on this topic

We have a requirement (R12) to be able to have the UNA locally
and an issue (5.18) on this.

Personally, I think that the fact that the UNA is NOT made,
makes things a lot harder to understand for "normal" non-DL
computer scientist, since databases and logic programming
do indeed make the assumption that two things with different
identifiers are not the same otherwise explicitly stated to
be so.

Hence, entailments made by OWL are quite often very surprising.
And models that one would intuitively consider as inconsistent are
consistent by entailments that make two things equivalent automagically.

Motivation I:
-------------
The reason I ask is, that I'm currently trying to come up with
a set of language primitives for OWL Lite that can be implemented
on top of SQL99-based databases (which are pretty much the defacto
standard today and support stratified datalog) and logic programming
systems (like Prolog / XSB).

While the domain-modelling part of RDFS and further property axioms
like transitivity, symmetry and inverse-of are "easy" to implement
everything that indeed states some constraints, for example functionality
of properties becomes problematic, since the default action is different
whether one has the UNA and not.

For example, with functional properties
one has to conclude that things are equivalent without UNA (and the model
would be consistent. It becomes inconsistent only when one says explicitly
that these things are
different), however with UNA the model would be inconsistent (and it becomes
consistent only when one says explicitly that things are the same).

The treatment of not having the UNA would have to be done entirely outside
of datalog for example with triggers and other mechanisms that "simulate"
the
inferences normaly made with a DL reasoner by adding or modifying ground
facts.
Technically, this is very ugly and very costly.

Motivation II:
--------------
The current guide document does not talk at all about having or not having
the UNA. However, we should say something about that and "prepare" people
on what they will get and what they can expect to happen.

Raphael
Received on Monday, 16 September 2002 04:21:52 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:52 GMT