W3C home > Mailing lists > Public > www-rdf-interest@w3.org > June 2004

Re: OWL DL restriction clarification, please

From: Benjamin Nowack <bnowack@appmosphere.com>
Date: Thu, 10 Jun 2004 19:08:17 +0200
To: Danny Ayers <danny666@virgilio.it>
Cc: www-rdf-interest@w3.org
Message-ID: <PM-EH.20040610190817.54399.1.1D@192.168.27.2>

On 10.06.2004 17:49:21, Danny Ayers wrote:
>Benjamin Nowack wrote:
>
>>On 10.06.2004 13:37:41, Danny Ayers wrote:
>>...
>>>
>>I think the specs are quite clear, the reference for example says:
>>[[
>>   "In OWL Lite and OWL DL, owl:Class must be used for all class 
>>   descriptions"
>>]]
>>  
>
>But that doesn't exclude the use of rdfs:Class as well, nor does it 
>explain *why*. In fact the very next point in the reference says:
I read, it _does_ exclude the use of rdfs:Class, for you would have
a class description that used rdfs:Class then.
The reason could be that you would need an additional inference
engine or a parser extension to detect if the used constructs/axioms
described an owl (DL) class. this way it's just easier to handle..
(hmm, although /me thinks ease of use has never been an argument
in the semantic web spec space, pls ignore this attempt to explain 
the *why* ;)
the *why* should be somewhere in webont minutes logs. I heard the 
rdfs:Class vs. owl:Class issue was a big one. but once they decided
to have owl:Class separate from rdfs:Class in owl DL, why should 
they re-open a potential can of worms by saying "rdfs:Class may 
still be used in DL onts, with the following restrictions: ...."?


>[[
>NOTE: |owl:Class| is defined as a subclass of |rdfs:Class|. The 
>rationale for having a separate OWL class construct lies in the 
>restrictions on OWL DL (and thus also on OWL Lite), which imply that not 
>all RDFS classes are legal OWL DL classes. In OWL Full these 
>restrictions do not exist and therefore |owl:Class| and |rdfs:Class| are 
>equivalent in OWL Full.
>]]

>...

>>Did you already try your example on e.g. protege's OWL 
>>plug-in? 
>>
>
>Not yet, I shall (I've got it set up with RACER here).
hm, I just remember, I'm not sure if the owl plug-in allows the
import of rdfs classes. but it's worth a try..

>...
>>I think one of the problems for deployed apps is that even if we had
>>DL/Lite-onts only, their combination could still result in an OWL Full
>>model. 
>>
>
>How? That's a pretty key question. My impression was that if you started 
>in the DL universe (i.e. DL or Lite) then you'd stay there.
don't know, e.g. if ont_1 defines a range for an owl:AnnotationProperty
of ont_2. (does that count? could be regarded as inconsistency. but 
there _are_ examples. I read it several times.)

>>So I guess the DL stuff is more for specialized apps that use
>>some kind of manually maintained/controlled set of onts.
>>
>>I *think* that many deployed vocabularies will be in the style
>>of the FOAF spec (rdfs:Classes with OWL information on top), maybe
>>offering a separate OWL DL version for those using DL reasoners.
>>(That's at least the scenario I'm optimizing my ont editor for.. ;)
>>
>
>I have seen a few in that style, but given that owl:Class is a subclass 
>of rdfs:Class, it seems to me that everything else being equal the OWL 
>version should be the first choice for new vocabularies, unless the 
>meta-definitions of RDFS are needed. So most data would be susceptible 
>to DL reasoning *and* whatever OWL Full/RDF reasoning was available.
time will tell. for the moment there are mainly rdfs vocabs and tools
out there who I guess are not going to replace their rdfs:Class with
owl:Class. (When we developed this image_regions vocab for the w3photo
project, I tried to promote the use of owl:Class, but we ended up
with rdfs:Classes quite quickly. And that's ok, not only for 
backwards compatibility reasons. OWL Full != awful ;)
I think it was Jim Hendler who said that the webont WG should have 
made it clearer that OWL Full is OWL, and OWL DL/Lite are just subsets.
But I also understand that people would like to use existing DL
reasoners.

>As an aside, I realise people in the SWBP WG don't want to tread on each 
>other's toes, but the issue of DL or Full is one which every 
>ontology/schema author will encounter. There is no need for them to 
>commit to advice pointing towards either OWL Full or DL (or Lite) 
>exclusively, but there should be information available on the 
>implications of any choice based on known facts. The mess won't clean 
>itself up, quite the opposite if people are implementing without full 
>knowledge of the relative strengths/merits of the alternatives.
I fully agree. 


>Cheers,
>Danny.

cheerio to italy.

--
Benjamin Nowack

Kruppstr. 100
45145 Essen, Germany
>
>
>
>-- 
>
>Raw
>http://dannyayers.com
>
Received on Thursday, 10 June 2004 13:08:28 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:07 GMT