ISSUE-29: Situation of owl:DataRange in OWL-Full

Hi!

It has been proposed [1] to close ISSUE-29 [2] as resolved by
 
  (1) identifying owl:DataRange with rdfs:Datatype

  (2) deprecating the use of owl:DataRange

To summarize my analysis below: AFAICS, point (1) would be consistent with the current specification of OWL-1.0-Full [3]. Also, at least from an OWL-1.0-Full perspective, point (2) would probably not lead to big problems (esp. if "deprecation" is understood in the sense of [4]).


I have checked all the tables in [3], which specify the characteristic semantic conditions of OWL-1.0-Full (in addition to the semantic conditions of RDFS).

A) Table "Characteristics of OWL classes, datatypes, and properties"

This table defines instances of class owl:DataRange to be instances of LV_I, which is the set of all data values in the OWL universe. Exactly the same semantic condition holds for rdfs:Datatype in the same table. So, there is no problem for identifying both classes from this perspective. 

However, there is an additional (informal) note in the entry of 'owl:DataRange', which says:

  "OWL dataranges are special kinds of datatypes."

I am not quite sure about the term "special" here. If it is *not* meant in the sense of "strict subset", then it doesn't oppose the identification of owl:DataRange with rdfs:Datatype. (And, if necessary, this note could be changed in OWL-1.1-Full, of course).

B) Table "Conditions concerning the parts of the OWL universe and syntactic categories":

This table does not contain an entry for owl:DataRange. This is somewhat strange IMO, and might be a bug in the Full spec. Nevertheless, it actually allows to add an entry for owl:DataRange to this table which says:

  If E is 	
    owl:DataRange 		
  then
    CEXT_I(S_I(E)) = IDC
    and IDC subsetOf C_I

This would then be exactly the same definition as for rdfs:Datatype, and so both properties would have the same property extention.

C) AFAICS, nothing in the Full spec says that 'owl:oneOf' has to be used in connection with 'owl:DataRange', when one defines enumerations of data values. This might be a bug in the Full spec, but it will make it easier to deprecate the use of 'owl:DataRange' in Full.

D) Table "Conditions on OWL vocabulary related to boolean combinations and sets":

This table contains the following entry for owl:oneOf constructs:

  IF AND ONLY IF
    C owl:oneOf l
  THEN
    C is a class, and
    l is a sequence of d1...dn over objects or datavalues, and
    the class extention of C is {d1...dn}

Here, the "back direction" is interesting: All classes which are /finite/ sets of datavalues will be expressable as oneOf statements. One could see this as a problem, *if* it was stated elsewhere that 'oneOf' statements define instances of owl:DataRange. In this case, it would not be clear whether owl:DataRange is a sub class of rdfs:Datatype or not. However, E) will show that we are out of danger:

E) Table "Comprehension conditions"

  IF
    there exists a sequence of datavalues x1,.,xn
  THEN
    there exists a y IN set IDC  // class extention of 'rdfs:Datatype' !
      with y owl:oneOf (x1...xn)

This can be seen as an addition to the back direction in D): It is said that for finite sequences of data values, we can create some 'owl:oneOf' construct, which actually denotes some instance of set IDC. And IDC is the class extention of rdfs:Datatype. Since every two such 'owl:oneOf' constructs for the same sequence of data values have to define the same class, we learn that the finite sets in D) actually define instances of rdfs:Datatype. So, again, identifying owl:DataRange with rdfs:Datatype is ok.

F) Table "Further conditions on owl:oneOf":

For triples "C owl:oneOf (d1 ... dn)", with data values d1...dn, it is specified that C has to be in the class IDC, which is actually the class extention of rdfs:Datatype. Hm, looks like redundant information, when compared wit D) and E) [FIXME!]. Anyway, again, identifying owl:DataRange with rdfs:Datatype is ok.


To conclude: Nothing in the current OWL-Full spec seems to speak against Peter's proposal to identify the class owl:DataRange with rdfs:Datatype. I would rather say that there are several hints which speek *pro* doing so. There is, however, a single point in the spec, which *could* be read as if owl:DataRange might be a strict sub class of rdfs:Datatype: It's the note cited in A). But this note is informal and somewhat vague, and I do not see any other support for having owl:DataRange really be a strict sub class of rdfs:Datatype. So Peter's proposal looks fine to me from the OWL-Full perspective.  


Cheers,
Michael

[1] <http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0017.html>
[2] <http://www.w3.org/2007/OWL/tracker/issues/29>
[3] <http://www.w3.org/TR/owl-semantics/rdfs.html#5.2>
[4] <http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0163.html>

--
Dipl.-Inform. Michael Schneider
FZI Forschungszentrum Informatik Karlsruhe
Abtl. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: Michael.Schneider@fzi.de
Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555

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 Regierungspräsidium Karlsruhe
Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus