Talk:Reason maintenance

From Wikipedia, the free encyclopedia
Jump to: navigation, search

Stallman and Sussman[edit]

The following was in the Implementations section:

TMSs were introduced by Doyle in 1979, while Stallman and Sussman's paper is from 1977. What McAllester says in the linked paper is that TMSs can be useful for dependency directed backtracking. Paolo Liberatore (Talk) 15:34, 17 September 2005 (UTC)

Dependency directed backtracking is simply a form of backtracking, many similar and more advanced algorithms are now used in constraint programming. Some of the original TMSs, including the one of Jon Doyle, make use of dependency directed backtracking as a means for resolving contradictions. I did not check this McAllester's paper now but what you say seems to be inaccurate or misleading at the very least. — Preceding unsigned comment added by Fraktalek (talkcontribs) 08:06, 30 March 2011 (UTC)


The articles said: "CML2, a Linux kernel configuration system developed by Eric Raymond, uses a truth maintenance system. However, CML2 has not been included in the standard distribution, or any distribution, of the Linux kernel"

However, the only source for this statement is [2]. The page describing CML2 in some depth [3] does not mention truth maintenance systems at all. I think that Raymond meant that CML2 can restore the satisfaction of constraints when some variabiles change, not that CML2 is actually implementing the TMSs as intended by Doyle, de Kleer, and the other ones cited in the "References" section. I have therefore removed CML2 as an implementation (being the only item there, I also removed the section). - Liberatore(T) 17:39, 2 January 2006 (UTC)

What Happened to TMS's?[edit]

Does anyone know why nothing has been published on TMS's in the last 15-ish years? Has some other concept replaced them? —Preceding unsigned comment added by (talk) 14:12, 31 October 2007 (UTC)

Assumption-based Truth Maintainance Systems were used in the early 90's as well in the context of some European Projects (ESPRIT) to investigate the use of Expert Systems in Fault Diagnosis and Control. TMS and its variants were eventually dropped from research agendas for two basic reasons: (a) their complexity is inherently exponential, and no clever dependency-directed backtracking or any other short-cut can eliminate this explosion, so such systems are in practice useless; if one wants to work with incomplete information (as is always the case in the real world) one must accept the fact that their conclusions may contain internal contradictions. (b) a lesser reason for the fading out of such systems research may have to do with the fact that the computer programming languages used to develop such systems were mainly LISP (in the US) and PROLOG (in Europe). Both these languages are niche languages and there exist few researchers fluent in these languages today —Preceding unsigned comment added by (talk) 12:45, 21 July 2010 (UTC)