Re: [TED] Action-188, ISSUE: production rule systems have "difficulty" with recursive rules in RIF Core

Gary Hallmark wrote:

> ...
> BTW, is "novar" a builtin, a constraint, or could never be in CORE 
> because it is too operational?

It is a metapredicate that can be expressed as a constraint (e.g., in
OSF, it is "X <: s" where X is a variable and s a minimal non-singleton
sort symbol; the semantics of  "<:" is "strictly subsumes"; e.g., in the
unsorted case such as Prolog, "nonvar(X)" is simply "X <: \top").

As you remarked during this past meeting, constraints are the perfect
abstraction mechanism to express builtins as long as order of resolution
does not matter. Even when order does matter, the discrimination between
what is an atom and what is a constraint can be exploited without a priori
"moving all constraints to one side" explicitly. (The "moving to one side"
is just symbolic to ease formal notation expressing the constrained
resolution rule as simply as posssible). Indeed, if important to the
operational semantics, the original order of subgoals in the RHS may be
preserved and the constrained resolution process made to process the
constraints in the order they are given (if such is needed to preserve
some order-sensitive operational semantics or effect).

-hak
-- 
Hassan Aït-Kaci
ILOG, Inc. - Product Division R&D
tel/fax: +1 (604) 930-5603 - email: hak @ ilog . com