W3C home > Mailing lists > Public > public-owl-wg@w3.org > May 2008

RE: ISSUE-125 (min1some): Min 1 QCR = someValuesFrom - Serialize as someValuesFrom?

From: Michael Schneider <schneid@fzi.de>
Date: Tue, 20 May 2008 22:00:41 +0200
Message-ID: <0EF30CAA69519C4CB91D01481AEA06A096AB1B@judith.fzi.de>
To: "OWL Working Group WG" <public-owl-wg@w3.org>
I strongly concur with Bijan's points, and want to add a few more.

First, I have to apologize to discuss this topic while the issue is still in 
"raised" state. But I cannot attend tomorrow's telco (travelling to Romania), 
so I am going to say, what I would say there, here.

It is intended that OWL provides different ways to express semantically 
equivalent things, because OWL is not only a reasoning formalism, but also a 
modeling language. That's why we now have owl:disjointUnion, which gives 
additional modeling power to OWL 2 in exchange for forward-compatibility, and 
without enhancing the semantic expressivity of the language.

OWL 1, btw., also contains a lot of syntactic sugar: owl:equivalentClass can 
be substituted by two rdfs:subClassOf axioms, which would bring certain OWL 1 
ontologies nearer to RDFS. Or there is owl:AllDifferent, or HasValue 
restrictions. Even owl:sameAs can be expressed by means of a nominal-based 
class assertions.

For the case of >=1-QCRs vs. SomeValues-restrictions: These are pretty 
different modeling tools, which just happen to be equivalent technically. For 
example, it might make sense, from a modeling perspective, to explicitly 
express [1..*] relationships between two classes, or even [0..*] 
relationships, although the latter would be redundant technically. Making 
these features illegal in OWL, and demanding to circumscribe them in a 
technically equivalent way, would not be what I want in such a case. Actually, 
this would be the situation of pre-OWL-2, where it was well known how to 
circumscribe QCRs. But people asked for QCRs often enough, anyway, probably 
not without a reason.

Even worse than disallowing >={0|1)-QCRs would it be to demand from the OWL 
tools do the transformation themselves. I just try to compare this with the 
strange situation where my Java programming IDE would rewrite all my generics, 
autoboxing, non-indexed loops, and all the other stuff which does not go into 
the bytecode eventually, just in order to make it more Java-1.0 compatible. I 
would certainly not use this IDE ever again. :) And then I try to imagine 
Topbraid Composer, which would have to serialize my >=1-QCR silently into a 
SomeValues-restriction. I expect this would probably lead to a lot of traffic 
in Holger's mailing list... :-/

Cheers,
Michael

>-----Original Message-----
>From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org]
>On Behalf Of Bijan Parsia
>Sent: Monday, May 19, 2008 10:37 PM
>To: OWL Working Group WG
>Subject: Re: ISSUE-125 (min1some): Min 1 QCR = someValuesFrom -
>Serialize as someValuesFrom?
>
>
>So:
>	1) intention hiding and non-roundtrippable; plus it frustrates the
>hell out of users when you silently change what they wrote
>	2) non-orthogonal; we need the general form in order to handle
>larger cardinalities anyway, so would have to impose a rather strange
>restriction
>	3) unnecessary; if users want to write their ontologies this way
>(so
>as to be compatible) then can easily do so, or postprocess.
>Furthermore, you could have a preprocessor before your old tool that
>did this, no need to build in this kind of strangeness into the base
>language.
>
>I propose closing this, with no change, on these grounds. I don't
>think we need to note the equivalence in the spec either (there are
>lots of equivalences...I don't see why this one is particularly
>interesting).
>
>Cheers,
>Bijan.



Received on Tuesday, 20 May 2008 20:01:23 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 20 May 2008 20:01:24 GMT