Re: [gnugo-devel] unused code

[Arend Bayer]

Gunnar wrote:

> If we're sufficiently happy with --experimental-connections I can
> remove the code and patterns it made superfluous.

Hmm. This one I would personally prefer to keep until
a) we have measured the performance impact of --experimental-connections
(this will give hints how much effort it is worth to improve the
performance of readconnect.c)
b) we are sure we have eliminated most of the weaknesses introduced by
--experimental-connections. (Currently replaying a NNGS game with
--standard-connections is a good way to find these.)

Also, I am yet to be completely convinced that --experimental-connections
is a already a clear gain (taking the speed tradeoff into account). All
we know is a favorable regression delta, a close win in a long twogtp
series you played. (Dan mentioned a rating increase on NNGS but that
happened at the same time as we released 3.3.10, and while GNU Go's
rating was still climbing fast.)

But you are probably better informed to judge this.

Arend