Re: [computer-go] KGS Tournaments: Uniqueness of players

[Gunnar Farnebäck <gunnar@xxxxxxxxxxxxxx>]

Nick wrote:
> I think that three are around 30 bots registered on KGS at present, and
> 23 of these are builds of GNU Go.  I do not want more than one of these 
> builds to enter any KGS Computer Go tournament.  Two non-GNUGo 
> programmers have told me that they would not be willing to enter an 
> event with more than one GNUGo competing.

For what it's worth, as a GNU Go developer I don't see much point in
multiple GNU Go entrants either. Sure, it could be useful for testing
various pieces of experimental code, but it's not exactly difficult to
get lots of games (against humans) anyway.

>      If someone takes GNU Go and adds a module to it that makes it play 
> better, I would not want to exclude the modified version.

How to handle derivative works is a tricky question.

> But I assume that such a module will soon get added to the
> "official" GNU Go anyway?

That depends on whether the person making the modifications is willing
to submit them for inclusion in GNU Go and to transfer copyrights to
the Free Software Foundation. The maintainers may or may not accept
the contribution but if it is indeed an improvement it's likely to
find it's way in sooner or later, possibly in a revised form.

> David Fotland has not added GTP to MFoG.  If some other person adds GTP 
> to MFoG, with David's permission, I would want it to be able to enter 
> KGS events.  If two other people both do this, I would only want one of 
> them to be able to enter - I think David should be able to decide which.

In general I think it's reasonable to only allow the author, or
somebody approved by the author, to enter a program.

> I would like SlugGo to be able to compete.  SlugGo contains GNU Go. 
> Should SlugGo and GNU Go both be allowed to compete in the same
> event?

This can be split into two questions. Should a derivative work be
accepted at all and when is a derivative work sufficiently unique to
be considered as a different program?

My suggestion is to accept a derivative work if

1. All copyright holders agree that they want it to enter the
   tournament and
2. the copyright holders can convince the tournament director that it
   is indeed unique enough to be interesting.

To clarify how this would work it should first be noted that GNU Go is
NOT in the public domain. It is copyright the Free Software Foundation
(FSF) and licensed under the GNU General Public License, which allows
certain actions (such as modification and redistribution) under
certain conditions (such as providing source code when the program is
distributed). A modified version of GNU Go (by someone who has not
transfered copyrights to FSF) would have a joint copyright between the
person doing the modifications and FSF. Thus point 1 would require the
FSF, in practice the GNU Go maintainers, to agree that it's desirable
to have it participate and point 2 would be very much up to the
tournament director.

A different example is SmartGo and possibly other programs which
utilize GoTools, which by point 1 would require agreement from Thomas
Wolf (which may or may not follow automatically from the license they
have).

As variations of 1 one may consider excluding pieces like the
communications code with no impact at all on the move generation or
accepting big enough modifications without requiring agreement from
other copyright holders. The former shouldn't be too much of a problem
but the latter would be difficult to define precisely.

Finally there's of course the possibility that somebody lies about the
derivative status of a program or uses someone elses code without
permission, but then we're in the realm of cheating.

> Maybe GNU Go would not want to compete in an event with the slow time 
> limits required by SlugGo?

GNU Go wouldn't have a problem with that.

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