RE: PRD review, part 1 and 2

Hi all, this is a comment from Patrick ALBERT, new to the group, but not
so new to Business Rules. I am actually a Fellow at ILOG, of which I was
a founder, and former CTO/VP R&D. for about 15 years, having started all
this by implementing an early object + Rules systems (KOOL, for
Knowledge and Object Oriented Language) about 25 years ago... 


About the rules semantic, I'd support Gary's statement below related to
how updates should impact rules activation: if we want clean semantic,
asserting a new object of a given class should activate all rules
relative to the class and superclasses of the new object: rules that
test the existence of such an object, and rules that test one or more


Updating a slot of an object should only activate the rules relative to
this slot. 


And, ideally, we might support a more declarative semantic making a
difference between equality and affectation. 


Affectation has the usual memory oriented semantic: a (possibly new)
value is placed in the slot of an object. It then triggers the relevant


Equality assert that the slot of a object has a given value. Three cases

 - the slot had no value: the new value is asserted and the rules that
test the slot are activated

 - the slot already had this value: nothing happens.

 - the slot had another value, and a contradiction is raised.



Such a definition allows for using rules for describing event triggered
actions -- basic production rules -- and declarative business logic.








> In my PR system, assign to a frame slot (aka a java bean property) is

> not the same as removing and asserting (or just re-asserting) the

> Assert will activate rules whose conditions reference *any* frame

> whereas assign will activate rules whose conditions reference the

> assigned slot.


Not sure I understand: can you give me a simple example?



Received on Tuesday, 19 February 2008 09:29:23 UTC