Re: Production Logic Programs

To the best of my understanding, Benjamin's Production Logic Programs 
are a very strong restriction of the (re)active rule paradigm of 
so-called production and event-condition-action rules.

The restriction lies in the assumptions made under point 2 on page 8 and 
point 3 onb page 9. This restriction makes it possible to declaratively 
specify (side-effect-prone) procedure calls in heads of rules, to 
collect (some) procedure calls during a processing of the rules, and to 
process the procedure calls only after processing of the rules. This 
surely is a semantically sound approach.

My doubt, however, is that this approach might suffice to express 
classical/expectable applications. At this point, I would advice not to 
design yet another rule formalism before application classes have been 
identified.


Another point: The claim that processing Production Logic Programs 
either by forward chaining or by backward chaining has no impact on 
their semantics does not seem to me to be correct. Forwards processing 
means deriving all possible conclusions regardless of (one or several) 
queries (so-called goals). With Production Logic Programs, forward 
chaining would in general generate more procedure calls (to be processed 
aftwerwards) than backward chaining from a goal.


Benjamin's email raises another thought: In my humble opinion, it would 
be preferable to investigate issues one by one. Meaningful steps might be:

1. what kind of (re)active rules are needed for what classes of 
applications -- this brings us back to  analyzing/classifying use cases,

2. what semantics are possible/desirable,

3. forms of reasoning are possible/desirable, eg backwards/forwards 
chaining, stratified or well-founded negatyion, that fulfill the 
selected semantics,

4. what syntactic constructs might both make rule programs easier to 
write and respect the chosen semantics

Francois

-- 
Francois Bry
http://rewerse.net  scientific coordinator
http://www.pms.ifi.lmu.de  head of unit

Received on Friday, 30 December 2005 12:18:34 UTC