Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: UCI versus Winboard

Author: Dieter Buerssner

Date: 10:58:16 08/21/02

Go up one level in this thread


On August 21, 2002 at 11:31:06, Robert Hyatt wrote:

>I don't consider "pondering" to be "basic stuff".

I agree with the pondering. However it was no prob for me at all. If UCI only
had a ponder command (without a suggested move to ponder) it would be solved.
Unfortunately, it doesn't.

>There are dozens of ways
>to do it, and that should be left to the engine completely...  Ditto for opening
>book, endgame tables, etc...

Where is a problem with book and EGTBs in the UCI protocol. You can use your
book just fine - as in WB. The handling of TBs is just better in UCI, because
not every engine has to setup things in their ini files, you get the info you
need from the GUI.

>> It leads to more stability IMO. Actually the
>>"control" is no big deal at all. If the engine has provided a pondermove, the
>>gui WILL instruct it to ponder that move. _Always_. What's the problem with
>>that?
>
>Just think about it for a minute.  We already have problems with auto232
>matches and strange things going on.  I'm not about to let some foreign
>piece of software control what/when my engine does things.  That is for my
>engine to decide, on its own.

I think, it is clear, that if you do not trust the GUI, there is no way. You
have to trust WB as well ...

>You say that like winboard is not robust.  Which I don't understand.  It has
>been robust enough for me to play nearly a million games on ICC/FICS/chess.net
>with no problems of any kind.  Not to mention playing matches for testing on
>my local machine.

I have found small bugs in my WB implementation after 2 years. I am not alone -
see for example Jon Dart's post. The comparision with Crafty is perhaps not fair
- WB has some code especially for Crafty. The common engine writer has to adapt
to it (for example how the analyze output is parsed when Crafty shows book
moves. This will for example make engines not work, that show at the end of the
PV a number in parentheses. Note, that the format of the PV is not defined in
the WB protocol).
>>>- UCI offers more information about the search
>>
>>> How?  The engine can tell the operator _anything_ it wants.  You can't provide
>>> more information than that.
>>
>>Yes, but I want it "live" on-screen in a windows environment, not in a
>>debug-file.
>
>Check out "telluser"...

I think, telluser is more or less useless for most of those things. I don't know
of any WB-GUI, that puts telluser messages into some (scrollable) window.
In UCI there are not only the free form messages (ala telluser) but also some
predefined things, like hash filling, CPU usage, tb hits, which are at least in
Shredder 5 displayed continiously at some point of the screen. It is no big
deal, but it makes things look a bit more interesting. There is no need to send
this info.

Other things, that UCI has: It can restrict the number of moves, which should be
searched. This can have various advantages while analyzing. It can for example
be used as an efficient "poor man's Multi PV mode" (it does not need a partitial
minimax search at the root). Also it has MultiPV mode (I don't support it - the
engines are free to not support it).

Some other disadvantages: It misses a new game and a result/end of game command.

As others, I think it will be easier to write support for UCI than for Winboard.
The main reason is, that it does not have all the "modes" of WB. Are we in force
mode? In analyze mode - but beaware, and accept moves and setboard during this
mode, etc. This, IMO, will also make it much easier to get it rather bugfree,
fast.

Regards,
Dieter



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.