Re: [RIF] Critical Success Factors for Goal 1

Hi Evan,

My replies are interspersed below

ewallace@cme.nist.gov wrote:

>
>
> Allen Ginsberg wrote:
>
>
> >Requirement:
> >       The RIF standard presupposes that given sets of
> >       rule-language families are provided.
> >       Initially, these sets may be specified purely
> >       extensionally (i.e., by enumerating the
> >       specific rule-langugage members).  Also, it is possible
> >       that intially (in phase 1 of
> >       the RIF activity) that only one such set is provided.
>
> What is the requirement here and on what/whom?  Is it that the
> RIF specification must define the rule-language families that
> it supports?
>
Good point - I was a little too quick on the old copy-and-paste.  As the 
"requirement" states, this is really a presupposition.
As I think of it now, I would actually say that it is a sub-CSF of CSF-1 
for the RIF to come up with meaningful rule-language families.

What makes a "meaningful" family?  There are a number of approaches. 
With the RIFRAF the WG seems to trying to use a top-down
approach to define rule-language families.  Back in December Hassan 
argued for a bottom-up approach 
(http://www.w3.org/2005/rules/wg/wiki/BottomUpClassification).   Earlier 
still the group compiled a list of rule systems 
(http://www.w3.org/2005/rules/wg/wiki/List_of_Rule_Systems).  

However it is done, and whatever technical terminology we come up with 
to express the idea, e.g., various well-defined "extensions" to the 
"core" RIF,
the end result, from a high-level point of view, will involve sets of 
similar rule-languages for which the RIF will hopefully make 
rule-exchange a reality.
I think the goals and CSFs of the RIF should be expressed using a 
high-level vocabulary.

>
> >Requirement:
> >       For each rule-language family (that the RIF supports)
> >       the RIF must provide a vendor/platform-independent
> >       canonical format for representing rules in that family.
>
> Why must the format be canonical?  An alternative would allow
> various representation forms for the same rule content as
> long as the meaning was unambiguous (as in OWL and RDF)?
> I don't object to this added constraint, but it should be
> explicitly justified.
>
> -Evan
>
I don't disagree that.  Perhaps "canonical" is not the best word to use, 
but as far as I can tell it is possible to have more than one canonical 
form for the same thing.
So, the requirement could be rephrased as "provide at least one 
canonical format...."

Thanks for your feedback,

Allen

Received on Friday, 19 May 2006 15:30:53 UTC