W3C home > Mailing lists > Public > public-owl-wg@w3.org > August 2009

Re: owl:Thing in RL profile?

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Fri, 14 Aug 2009 09:32:24 +0100
Cc: Ian Horrocks <ian.horrocks@comlab.ox.ac.uk>, Boris Motik <boris.motik@comlab.ox.ac.uk>, W3C OWL Working Group <public-owl-wg@w3.org>
Message-Id: <C601BCF2-8746-4338-93C8-E3E15674E87A@cs.man.ac.uk>
To: Alan Ruttenberg <alanruttenberg@gmail.com>
On 14 Aug 2009, at 06:32, Alan Ruttenberg wrote:

> On Thu, Aug 13, 2009 at 5:05 PM, Bijan Parsia<bparsia@cs.man.ac.uk>  
> wrote:
>> On 13 Aug 2009, at 10:31, Ian Horrocks wrote:
>>> I agree with what Boris says. There are several features that we  
>>> excluded
>>> from RL on the grounds that they would hamper practical  
>>> implementation. I
>>> believe that this was one of them. The argument goes that adding  
>>> owl:Thing
>>> to the profile would require a rule that adds the relevant type  
>>> triple for
>>> every individual. Of course there is nothing to prevent RL  
>>> implementations
>>> from dealing correctly with owl:Thing, but they are not required  
>>> to do so in
>>> order to be conformant.
>> Plus, frankly, owl:Thing instantiation entailments (or assertions)  
>> are
>> silly, given their tautologous nature.
> They are silly in exactly the same way the
> owl:Top(Data|Object)Property entailments are silly, no?

Please note that I said *instantiation* entailments. E.g.,  
ClassAssertions(Thing, x). Not *all* entailments (e.g., where thing is  
subclass something else). And C subClassOf Thing is "silly" as well  
(not logically silly, but not a thing to focus on in a query).

In the RDFS rules, every implementor has to recognize that  
instantiating rdfs:Resource is a non-starter. Every UI or Query  
designer *should* recognize that returning such results or displaying  
them tabularly is harmful, by and large. In a tree view, it's of  
course, reasonable to put all individuals without a more specific  
class into owl:Thing.

But for instantiation assertions (i.e., that two individuals are  
related by the TopProperty), yeah, they are similarly silly and rather  
worse computationally (cross product). They are also not in RL:
"""Furthermore, OWL 2 RL does not support the predefined object and  
data properties owl:topObjectProperty, owl:bottomObjectProperty,  
owl:topDataProperty, and owl:bottomDataProperty."""

Received on Friday, 14 August 2009 09:05:49 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:42:13 UTC