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

RE: proposed resolution of Qualified Restrictions

From: Ian Horrocks <horrocks@cs.man.ac.uk>
Date: Mon, 29 Apr 2002 10:41:14 +0100 (BST)
Message-ID: <15565.5434.924094.555077@excalibur.cs.man.ac.uk>
To: <www-webont-wg@w3.org>
On April 25, Jeremy Carroll writes:
> 
> >
> > Can this be confirmed i.e. does this mean that there is _effectively_ no
> > difference between the qualified and unqualified restrictions?
> 
> No no, that wasn't my point.
> 
> My understanding is that the qualified ones are in DAML+OIL because they
> were free to implement. i.e. the additional cost of implementing them over
> the unqualified ones was trivial.
> 
> I think that they do add real expressiveness to the language.
> The case against them is that even if that expressiveness is free to
> implement, it costs learners, documentors, ontology designers etc.
> Given that the particular expressiveness is close to useless, then a
> cost-benefit analysis suggests it goes.

Proving that one language is strictly more expressive than another can
be tricky, and although I believe that qualified number restrictions
really do add to the expressive power of OWL, I do not recall seeing
any proof that they are more expressive than simple number
restrictions. They certainly do, however, make it easy to express some
things (like Mules) that are very difficult to express without them.

QNRs were included in DAML+OIL because of their usefulness in some
situations (like mules), and because supporting them is relatively
easy from an implementation point of view - in fact they are just a
generalisation of toClass and hasClass restrictions. In particular,
(hasClass P C) is obviously equivalent to (minCardinalityQ P 1 C).

Regards, Ian

> 
> Jeremy
> 
> 
Received on Monday, 29 April 2002 05:48:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:49 GMT