Re: computer-go: Is the Go Modem Protocol used in actual tournam ents ?

[Daniel Bump <bump@xxxxxxxxxxxxxxxxxx>]

Nick Wedd wrote:

> I can confirm that various programs which implement GMP do so partially
> and idiosyncratically.  When two such programs play each other, they may
> or may not be able to use GMP successfully.  If not, it is unclear which
> one is to blame.
> 
> A robust and generally used GMP implementation would be a boon for
> tournament organisers.

And, I would add, for progrmmers!

A start would be to rewrite the docs. The GMP *is* documented but
the existing doc could use clarification, it seems to me. SGF
has greatly benefited by the attention it's gotten from
standardization-minded folks. (http://www.red-bean.com/sgf/)

Also, as others have mentioned at various times, a standardized text 
based protocol would be good. GNU go has a start of a text
based protocol in it, which needs development. Could this be
the basis of a standardized text protocol?

Francois Grieu wrote:

> Unfortunately we have at least two different sources floating
> around (not counting CGFCOM.DLL). One comes with the definition
> of the standard by David Fotland, the other is by William Shubert
> and is in goDummy-1.0.4, available at
> <http://www.igoweb.org/~wms/comp/cgoban/index.html>
> or nearly equivalently the development version of GnuGo
> <http://www.lysator.liu.se/~gunnar/gnugo/devel.html>

This is only unfortunate to the extent that Bill Shubert's
version doesn't faithfully implement the protocol. I'm
aware of one defect in Bill Shubert's file, which is that
it doesn't implement komi. In 1999 GNU Go played in the US
Go Congress (the only tournament it's ever participated in
to date) and it was successfully able to communicate with
both Many Faces and SmartGo over a serial cable using the
protocol.

Dan