Re: RIF UCR REVIEW

Hi,

> 
> Dave Reynolds wrote:
> > Adrian Giurca wrote:
> >
> >>    * I believe that the UCR document needs to contain rule 
> examples in
> >>      different rule languages and not just a natural language rule
> >>      text. This will help to better understanding the RIF 
> requirements.>
> > I thought we made an explicit decision not to do this. I think 
> this 
> > was to avoid readers having to understand different rule 
> languages, to 
> > avoid setting incorrect expectations on what rule languages might 
> be 
> > mapped into RIF and to reinforce that the rules are simplified 
> > examples only.
> I was not aware about that. However, I guess it is not simple to 
> derive 
> requirements if examples are not expressed in concrete languages.

We have discussed and decided upon the process of gathering requirements
within the working group. A number of steps have been followed for
determining the requirements stated in the 2nd WD UCR, which are mainly
Phase I requirements. The authors of the proposed use cases have stated
explicit requirements their use cases pose on RIF. I have also analyzed
the (explicit and implicit) requirements coming from the use cases.
Moreover, each WG participant has had the possibility to propose the
requirements he or she considered important for RIF.

As you've noticed, we just started to gather Phase II requirements now
and more or less the same approach is considered.
 
> Each 
> language has its own rule representation and requirements. Their 
> main 
> requirements must be captured by RIF. 

IMO, the followed approach for gathering requirements is a well suited
one. I don't think that concrete rule language should directly pose
requirements on RIF. For this, we have the requirement on rule language
coverage, which acts as an umbrella requirement for the ones coming from
RIFRAF.

>For example rules expressed 
> in 
> Prolog must conform to some requirements (see for example  
> anonymous 
> variable from Core) while production rules like JBoss Rules have 
> other 
> requirements. I see here three categories of languages:
> Classical AI rule languages: Prolog, L-Logic, Jess etc
> Semantic Web Rule Languages: SWRL, Jena 2, RuleML etc
> Production rule systems : JBoss Rules, etc
> Each of them has specific requirements so the RIF must somehow 
> identify 
> the common part.

As already mentioned, such kinds of requirements will be derived from
RIFRAF.


Regards,
Paula