- From: Gary Hallmark <gary.hallmark@oracle.com>
- Date: Mon, 22 Sep 2008 13:01:55 -0700
- To: rif WG <public-rif-wg@w3.org>
As shown in [1] and [2], forward (backward) chaining systems may have trouble if variables are bound to predicates/frames in the head (body) of a rule. So for Core, we have to decide whether one style is preferred to the other. I know that in most PR systems, all variables must be bound to predicates or frames in the condition. We need a "PRD Core" that defines a subset of PRD where all variables are bound to predicates or frames in the condition. This will allow a PRD translator to be implemented without the burden of equation solving for the frame/predicate variables in body should you receive a rule in which they are bound in the head Is this too onerous for Core? I imagine that a "BLD Core" would want just the reverse: all variables are bound to predicates or frames in the head. [1] http://www.w3.org/2005/rules/wiki/Chaining_strategy_numeric-subtract_1 [2] http://www.w3.org/2005/rules/wiki/Chaining_strategy_numeric-add_1
Received on Monday, 22 September 2008 20:03:42 UTC