Re: On production rules and phase I&II

Dear All,

I feel very, very uneasy at attempts to specify a declarative semantics
for reactive (or production) rules in terms of declarative (e.g. Horn)
rule counterparts. Indeed, this is possible -- but under very strong
assumptions, like no negation, that are unrealistic in practice.
Furthermore, I believe this is useless for applications.

Reactive rules (including production rules) are inherently imperative
because they specify state transitions. Admitteldy, while in standard
imperative languages states refer to variables, in reactive rule
languages states refer to a "database". But this distinction does not
make reactive rule language more declarative or less imperative.

I suggest to keep bothy fragments of RIF for what they are:
- a declarative fragment with a declarative semantics
- an imperative fragment (that of reactive rules) with an imperative
semantics.

This is what application need. A declarative semantics for reactive (or
production) rules might be a good subject for academic research. Or
might well not be -- my own guess.

Regards,

François

Received on Wednesday, 8 March 2006 11:51:00 UTC