- From: Michael Kifer <kifer@cs.sunysb.edu>
- Date: Tue, 19 Dec 2006 06:31:41 -0500
- To: Francis McCabe <frankmccabe@sandsoft.com>
- Cc: Gary Hallmark <gary.hallmark@oracle.com>, public-rif-wg@w3.org
Francis McCabe <frankmccabe@sandsoft.com> wrote: > > The issue with running out of memory speaks to a couple of core > differences between PRR-based systems and LP-based systems. The most > obvious is, of course, the forward-chaining vs backward chaining. A > more subtle, but perhaps more significant difference is in the way > disjunction is handled. Many years ago I stumbled upon a CLIPS manual where it was described how to avoid computing the entire model. It was essentially a description of a rudimentary Magic Set transformation. The date on that manual was 1983 or 1984 - a year or two before the first magic set paper appeared. The point is that one can write bad rules bottom up or top down - doesn't matter. You can avoid computing the entire model, using smart techniques. That CLIPS manual author of > 20 years ago was definitely very smart. Too bad he didn't think that his trick was worth a publication :-) --michael
Received on Tuesday, 19 December 2006 11:34:28 UTC