[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [computer-go] how to use GTP in place of GMP



In message <XFMail.20040811233234.Hellwig.Geisse@xxxxxxxxxxxxxxxxx>, Hellwig.Geisse@xxxxxxxxxxxxxxxxx writes
I think a viable solution for the practical problem
how to specify the communication of Go programs for
a tournament could be like this:

1. The GTP protocol is not affected in any way.
  It remains exactly as is, without specifying
  the transport mechanism.

2. The participants are asked to supply programs
  which connect via TCP/IP, address xx.yy.zz.vv,
  port ww, and speak GTP over this connection.
  (It might be necessary to specify the physical
  connection also, in case the computer on which
  the participant's Go program is running is not
  supplied by the organizer.)

3. We, the developers who want to further GTP, will
  supply (at least) two packages, one for Windows
  and one for Unix, containing reference implementations
  of a referee (well, not really a referee, but merely
  what we used to call a controller). I think that these
  packages are not much more than one of the twogtp
  programs, 'glue' for connecting them to TCP/IP, together
  with instructions how to install and use them.

4. The organizer of a tournament then only has to supply
  one computer per pair of Go players with one of the
  packages installed on it. This computer must of course
  be connected to a LAN.

Nick, would this solve or at least reduce the
problems you saw for organizing a computer Go
tournament?
If I have understood this right, I definitely like it. It sounds as if it would remove all the problems that I have complained of. At least, it would remove them from the tournament organiser. I can't judge how hard it would be for entrants to supply programs as specified in (2).

Is a "reference implementation" an actual program, which I merely have to run? I hope so.

I would like to suggest, if I may, a few features which this program might have. It would write, for each game, a file containing an SGF record of the game. This file would contain time information in BL and WL fields. The file would be written after each move; so that after a lightning strike or other accident, the game could be resumed.

I have one reservation. So far as I know, there is currently no program which thinks (uses the CPU etc.) in its opponent's time. But this ought to be allowed, if anyone implements it. This would preclude having pairs of players cohabit on one machine.

Nick
--
Nick Wedd nick@xxxxxxxxxxxxxxxxx
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/