RE: Fragments discussion, continued

On Tue, 19 Feb 2008, Boris Motik wrote:
>
> Hello,

Hi Boris,

> I agree with this proposal in most of its main points; however, I do
> believe that there is a strong case for adding DLP to the list of the
> official fragments. Here is why I believe that this is so:

What does official mean? I understand that all of the fragments would
go into some ("official"?) document, but only some of them would be
rec track. For the remainder of my reply, I assume that official==rec.

> - DLP is distinguished by the fact that you can do reasoning by
> looking only at the individuals explicitly introduced in the
> ontology. Many simple implementations of OWL will actually implement
> precisely this fragment.

I thought this is precisely the motivation of having OWL Prime on
rec. I cannot see why, additionally, we would want to have DLP on rec.

Also: When you talk about "many simple implementations", do you mean
OWL DL or OWL Full/RDF?

> - DLP seems to be the perfect vehicle for explaining the
> relationship between OWL Prime and the DL fragments. 
>
> The reasoning in
> OWL Prime is also done by just considering the individuals explicitly
> present in the ABox, so there is a natural correspondence between it
> and DLP.

I agree. Only: why does this have to be in a rec document? We will have
an additional, non-rec document on fragments. It should of course 
include DLP, and IMHO it would be just the appropriate place for 
explaining the relationship you mention.

> In fact, in view of my latter statement, we might think of having an
> OWL Prime/DLP fragment: OWL Prime would be the RDF face and DLP would
> be the DL face of the same fragment. Please note that I do not
> propose to use these names; the name "DLP" as such is not important
> to me. I just believe that, even on the DL side, there is a strong
> case for having a fragment such as DLP.

Sorry, I don't see the strong case. Being an "asymmetric" fragment
(different constructors allowed on the left- and right-hand sides of
GCIs), I don't find DLP very appealing as a fragment of OWL DL. It
didn't seem to receive much uptake since its invention either (correct
me here if I overlook something). There is no implementation, and no
ontology formulated in it. In fact, not long ago many people of this
WG agreed to drop DLP altogether. At that time, the connection to OWL
Prime was unknown, but I fail to spot the advantage of having both OWL
Prime and DLP on rec, when we could have OWL Prime on rec (which has
Oracle support) and clarify in an auxiliary document how it relates to
OWL DL (via DLP).

greetings,
 		Carsten

--
*      Carsten Lutz, Institut f"ur Theoretische Informatik, TU Dresden       *
*     Office phone:++49 351 46339171   mailto:lutz@tcs.inf.tu-dresden.de     *

Received on Tuesday, 19 February 2008 20:15:16 UTC