Part II: Response to Peter F. Patel-Schneider [RE: review of RDF-Based Semantics]

Hi, Peter!

In this mail I will answer your comment concerning owl:NamedIndividual and Key axioms.

Peter F. Patel-Schneider wrote on March 30, 2009:

>Technical Changes:
>
>1/ Align keys with treatment in direct semantics:
>   5.2:
>    owl:NamedIndividual	\in IC 	\subseteq IR
>  5.2:
>    Remove the "Informative Note".
>  5.14:
>    Add x,y \in NamedIndividual after the "if" in the RHS of Table 5.14

There are actually two issues here:

(1) Should the class extension of owl:NamedIndividuals be made a /sub/set of the domain? (Currently, it is defined to equal the domain.)

(2) Should the semantic condition for Key axioms be adjusted the way you propose?

I will treat these two issues separately.

ad (1): 

I remember that we had this same discussion earlier when you reviewed the document before publication as FPWD. At that time I argued in favor of having the class extension of owl:NamedIndividual being equal to the domain, although it may seem counter intuitive, since there were some technical reasons for this. See:

  <http://lists.w3.org/Archives/Public/public-owl-wg/2008Sep/0157.html>

However, it now turns out that the original technical reasons have basically gone (or at least weakened), since the relationship between the RDF-Based Semantics and the Direct Semantics is now measured in a somewhat different form than before. 

I also see no big trouble for actual usage outside the constraints of valid OWL 2 DL ontologies. If people do not care about authoring valid OWL 2 DL, then I consider it rather unlikely that these people will make much use of owl:NamedIndividual, in particular not in the way that they declare names as owl:NamedIndividual on the RHS of an entailment, without doing so on the left hand side. But if they really do, then the situation for (named) individuals is not different from that of other entity types, such as classes or properties: it may happen sometimes lead to an OWL Full non-entailment. This is something OWL Full folks (and their reasoners) will have to learn in general, and is not specific to owl:NamedIndividual.

I have therefore followed your suggestion to weaken the semantics of owl:NamedIndividual in the way you propose. I have also removed the "Informative Note", as you propose, since it is now wrong (more precisely: the last sentence becomes wrong; the rest of the note becomes redundant, since it served as a rational for the last sentence).

DIFF: <http://www.w3.org/2007/OWL/wiki/index.php?title=RDF-Based_Semantics&diff=21279&oldid=21274>
 
ad (2):

First, if x and y are required to be instances of NamedIndividual, as a "simulation" for the notion of being named in the ontology, then I think this has to be the case for the other variables as well, because the individuals are also required to be ISNAMED in the Direct Semantics version of keys axioms.

However, I believe that NamedIndividual is not an adequate substitution for the "ISNAMED" operator in the Direct Semantics. Even for valid OWL 2 DL ontologies it is not required in every case that a (named) individual IRI needs to be declared as an owl:NamedIndividual. So requiring that the variables are instances of owl:NamedIndividual would lead to a /weaker/ semantic condition compared to the Direct Semantics. 

But I see an even bigger problem with the practical use of Key axioms under the RDF-Based Semantics, in particular when people don't care about OWL 2 DL-ness of their ontologies.  If we would do the change as you propose, then people would be required to add owl:NamedIndividual declarations to names in order to apply the semantic condition of owl:hasKey axioms. For example, if one wants to use Keys with an existing triple store, one would first have to add owl:NamedIndividual typing triples to the resources, in order to make the axiom work. 
 
Further (and related to the previous point), this change would also have a side effect on the "Profiles" document. The OWL 2 RL rule for "owl:hasKey" would need to be changed as well, by requiring that "?x" and "?y" (and potentially the other variables, either) be of type owl:NamedIndividual. Otherwise, the OWL 2 RL rules would become unsound w.r.t. OWL 2 Full, because they would fire "without restriction" (even on bNodes, for example).

Based on these arguments, I prefer to *not* perform the change.
   
Best,
Michael

--
Dipl.-Inform. Michael Schneider
Research Scientist, Dept. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: michael.schneider@fzi.de
WWW  : http://www.fzi.de/michael.schneider
=======================================================================
FZI Forschungszentrum Informatik an der Universität Karlsruhe
Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
Tel.: +49-721-9654-0, Fax: +49-721-9654-959
Stiftung des bürgerlichen Rechts, Az 14-0563.1, RP Karlsruhe
Vorstand: Prof. Dr.-Ing. Rüdiger Dillmann, Dipl. Wi.-Ing. Michael Flor,
Prof. Dr. Dr. h.c. Wolffried Stucky, Prof. Dr. Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
=======================================================================