Author: Peter Schäfer
Date: 01:09:33 04/19/04
Go up one level in this thread
On April 17, 2004 at 14:08:06, Stefan Meyer-Kahlen wrote: >On April 16, 2004 at 15:17:17, Peter Fendrich wrote: > >>On April 16, 2004 at 12:26:11, Stefan Meyer-Kahlen wrote: >> >>> >>>Hello Peter, >>> >>> >>>On April 16, 2004 at 12:14:37, Peter Schäfer wrote: >>>> >>>>Hello Stefan, >>>> >>>>Is there a chance that we will see further improvements to the UCI protocol ? >>> >>> >>>Yes, there will be further improvements, but we have just released a new version >>>so don't expect it to happen next week. >>> >>> >>>>I guess these issues have been discussed before a lot, just to mention a few: >>>> >>>>- let the engine decide about resignation and draw. >>>> This is even more interesting if the engine can take the opponent's ELO >>>> into account (e.g. don't offer draws to weak opponents). >>> >>> >>>We have discussed this in Paderborn and we sort of agreed that this would wreck >>>the stateless design of the protocol. We integrated AnalysisMode and ucinewgame >>>but we didn't want to go any further. >> >> >>Why not another info tags: "resign" meaning "I want to resign" and "draw" >>meaning "I want a draw". Then it's up to the GUI (the user) to handle it as >>desired. > > >In "pure" UCI the engine doesn't know (and doesn't need to know) that it is >playing a game. The engine just gets positions to search. Actually the engine >doesn't know what a game of chess is. Because of this the engine can't resign a >game or offer draws in a game. > >Now we have added the "AnalyseMode" command, so the engine does at least know >whether the positon to search belongs to a game or is just an analysis. Also >with "ucinewgame" the engine knows to which "game" (actually it need not be a >game for the engine, just that after ucinewgame the engine is searching on >something different) a position belongs. We did this because of public demand >and not because we really liked it. > >Those two changes are a step in the direction to explain the engine what a game >is. If you want the engine to resign something you have to tell the engine what >to resign. We didn't want to go that far. It is just a design or philosophical >decision. > Hello Stafan, this seems to be one more point in favor of a "stateful" protocol. I don't want to start a discussion about it, because such a descision needs careful consideration (and you certainly have discussed it before). I just want to point out that many engines "de facto" already use a kind of stateful approach, because it is common practice to transfer the complete move list with every position. It is not an elegant solution, but it demonstrates that there is a real demand for a stateful protocol. Best Wishes, Peter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.