Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!ux1.cso.uiuc.edu!ux1.cso.uiuc.edu!osiris.cso.uiuc.edu!goldfain
From: goldfain@osiris.cso.uiuc.edu
Newsgroups: comp.lang.prolog
Subject: Re: different Prologs?
Message-ID: <7000015@osiris.cso.uiuc.edu>
Date: 17 Oct 89 09:40:15 GMT
References: <111000002@uxa.cso.uiuc.edu>
Lines: 27
Nf-ID: #R:uxa.cso.uiuc.edu:111000002:osiris.cso.uiuc.edu:7000015:000:1504
Nf-From: osiris.cso.uiuc.edu!goldfain    Oct 16 21:08:00 1989


] /* Oct 11, 1989   ok@cs.mu.oz.au writes in comp.lang.prolog */
] ...
] There's another sense of "reasonable"; intelligent backtracking ranges from
] outrageously costly to unpleasantly costly.  There are affordable versions
] of intelligent backtracking around, but they become affordable by sacrificing
] much of the intelligence.  You can avoid repeated solutions to a goal _this_
] simple, but the problem is going to come right back in more complicated
] situations.  Another method is to use extension tables (applicable only to
] :- pure predicates), which have their own problems.
] ...
] As they say in the satellite business, there's no such thing as a free
] launch.  Prolog strikes a balance between cost and power; it's fast and
] stupid rather than slow and smart.
/* End of text from comp.lang.prolog */

Hmmm... This is one other place where one could explore parallelism: have
separate processors hunting for the best or simplest backtrack scheme.
This sounds vaguely like it might be a psychologically plausible model.  I
don't know what kind of a nightmare this would be to implement.

I think the general approach to parallelism in Prolog is somewhat incompatible
with this idea - all promising approaches are being explored, so an
intelligent backtrack would never be involved.  But I see no proof that the
idea doesn't have unique benefits in some problem areas, so I can't dismiss it
as a pointless investigation.
                                                  - Mark Goldfain