Re: [TED] Question on constraints

Hassan Aït-Kaci wrote:
> 
> I am not sure I understand exacty what your point is.

That may be because I completely misunderstood what your point was...

Let me try to make my point clearer...

Initially, I understood that the benefit of the CFP formalism was to
isolate the part of a rule's condition that was really constraining the
instantiation of the rule, that is, that specified (or constrained) the
valuation domains for the rule variables (the variables that are free in
both the antecedent and the consequent of the rule). Those constraints,
bearing on the rule variables only, are naturally defined at the same
level where the rule variables are quantified, that is, as a component
of a rule (along with variable quantifications, the antecedent
expression and the consequent expression).

That made sense to me

Now, you seem to say that the benefit is really to abstract the data
model from the rule. I do not completely understand how this trick 
works, but, anyway, it seems to me that, in that case, all the arguments 
in the antecedent expression and all arguments in the conseuquent are 
concerned, rule variables or not. But the constraints that bear on 
arguments that are local to the antecedent, resp. the consequent, should 
be associated, in the meta-model, with said antecedent, resp. 
consequent, not the rule. And , thus, we should have constraints 
associated with the rule (for the rule variables), the antecedent, and 
the consequent, not only the rule.

Does my question make any more snese, after the explanation?

Christian

> Christian de Sainte Marie wrote:
> 
>> Hassan Aït-Kaci wrote:
>>
>>>
>>> (3) other data models may be accommodated using this "data model as
>>>     constraint system" paradigm and conjugated with definite clauses
>>>     using the CLP scheme in order to obtain Horn-like rule systems
>>>     over varied data types (e.g., Java, C#, or C++ classes/objects).
>>>
>>
>> We agreed that RIF would specify a common expression language from 
>> which the various RIF dialects will draw (by restricting which 
>> expressions are allowed in the LHS/condition, RHS/conclusion, and 
>> constraint expressions [1]).
>>
>> [1] E.g., CORE will probably allow only a conjunction of atoms in the 
>> body and a single atom in the head.
>>
>> Don't we need to associate constraints to said language expressions, 
>> instead of at the rule level only (as we discussed and decided at 
>> F2F4), if we want that good property of data model abstraction brought 
>> by the constraint paradigm to apply to the ground arguments and 
>> locally bound variables in the expressions?
>>
>> If not, how comes that we do not need to abstract the data model in 
>> order to agree on the semantics of expressions in the expression 
>> language, but we need it for rules (or that the abstraction can be 
>> implicit at the expression level but not at the rule level)?
>>
>> Christian
>>
>>
>>
> 
>

Received on Friday, 1 December 2006 14:23:42 UTC