Re: ISSUE-43: Subclass relationship ## in RIF-BLD

> 
> Thanks Michael,
> 
> > Objections have been raised to the inclusion of the subclass relationship,
> > a##b, in RIF-BLD on the grounds that it duplicates rdfs:subclassOf.
> 
> I think it would make sense to generalise the objection to:
> <<
> RIF including subclass (and other similar metamodel-type relationships?)
> duplicates the data model specifications used to provide data for RIF
> rules. >>

The question is what is a data model. Many KR people consider Isa hierarchy
to be as much part of the application as of the data model. And it really
is -- unlike saying that the type of a property abc is foobar.


> I have to say I am neutral on the topic - with maybe a slight preference for keeping it out of RIF BLD on the basis that 
> - BLD is (AFAIK) meant to be a "common dialect subset", 

It does not shape up this way. BLD has a lot of stuff that most
languages do not support. Frames, to name just one :-)

> - subclass relationships are simply a special type of relationship test (eg why not consider PartOf relationships?) (or maybe relationship assertion),

PartOf does not have logical semantics. It is just a UML convention.

On the other hand, "subproperty" is a reasonable ting to add. However, this
is by far not as common as subclass. For instance neither Java, nor SQL
(incl. SQL/3), nor Entity-Relationship model (of the standard variety --
there are extensions that include everything under the sun), nor C++, nor
Smalltalk, nor F-logic have a builtin construct for subproperties.

> - in Prod Rules we "generally generalize" to "memberOf" ie X is a memberOf parts of Y or subclasses of Y (etc). Due to most PR languages using Java as their 
> base object model, and with multiple inheritance being passé in that world, OO / frame relationship tests tend to be rare* (to say the least), and assertions 


But this is RIF-BLD, not RIF-PRD! In RIF-PRD you might make a different
decision.


> - well...  
> 
> But I understand this issue is of much more importance to the KR folk, which is why I am officially neutral (for now!).

Good :-)


> Paul Vincent
> TIBCO | ETG/Business Rules 
> 
> PS: * You could argue that logically, PR are often testing for "subclass" - its just the the object model implementation / used for writing the rules is not 
> defined in a pure (multiple inheritance) OO fashion, so the data model is "flattened" compared to other OO models ...

Maybe. But PRD is a different beast than BLD, and we are discussion BLD.


	cheers
	  --michael  


> 
> > -----Original Message-----
> > From: public-rif-wg-request@w3.org [mailto:public-rif-wg-request@w3.org]
> > On Behalf Of Rule Interchange Format (RIF) Working Group Issue Tracker
> > Sent: 16 October 2007 17:10
> > To: public-rif-wg@w3.org
> > Subject: ISSUE-43: Subclass relationship ## in RIF-BLD
> > 
> > 
> > 
> > ISSUE-43: Subclass relationship ## in RIF-BLD
> > 
> > http://www.w3.org/2005/rules/wg/track/issues/43
> > 
> > Raised by: Michael Kifer
> > On product: Architecture
> > 
> > Objections have been raised to the inclusion of the subclass relationship,
> > a##b, in RIF-BLD on the grounds that it duplicates rdfs:subclassOf.
> > 
> > Defenders of this relationship state that:
> > 
> > 1. Subclass relationship is a common and very basic concept of any
> >    object-oriented/frame representation, and frame representation
> > requirement
> >    is in the Charter.
> > 
> > 2. The rdfs:subclassOf relationship is not a standard subclassOf
> > relationship.
> >    Using it instead of ## introduces additional axioms into the semantics,
> >    which are not supported by standard object-oriented languages.
> > 
> > 3. Not including ## in the language means that systems like FLORA-2,
> >    Ontobroker, FLORID, etc. must invent a new dialect to exchange their
> >    Horn subsets just because ## is not included.
> > 
> > 4. Excluding this construct precludes us from stating simple things like
> >    bar##foo (i.e., bar is a subclass of foo) without carrying the baggage
> >    of the additional axioms of rdfs:subclassOf.
> > 
> > 5. ## does not preclude the use of rdfs:subclassOf for languages that want
> >    to use RDFS' notion of subclass.
> > 
> > 
> > 
> 
> 
> 
>

Received on Tuesday, 16 October 2007 19:18:36 UTC