[computer-go] GTP and handicap

[Don Dailey]

Yes,  I'm not really fussy about this,  it's just that there should be a
simple way to fully specify any ruleset - via GTP.   

What I was seeking was a more universal approach - not a GTP specific
approach.

- Don



On Thu, 2006-09-07 at 23:11 +0200, Andrés Domínguez wrote:
> 2006/9/7, Don Dailey <drdailey at cox.net>:
> > On Thu, 2006-09-07 at 22:01 +0200, Rémi Coulom wrote:
> > > By the way, GTP needs a way to specify the rules of the game.
> >
> > Yes, I agree with this.   There should be a very clean way to specify
> > the rules of the game in a very concise and unambiguous way.
> 
> Me too.
> 
> > It could be coded up in a string, where each character of the string has
> > some meaning, such as whether suicide is allowed, which version of KO is
> > used, scoring system used, etc.   Then it could be expressed very
> > conveniently in a short string.
> 
> I disagree. I think compact is bad (and of course is not GTP's style).
> One rule, one command with the argument. Ex.
> 
> scoring_rule japanese
>                   chinese
> 
> ko_rule simple
>            positional_superko
>            situational_superko
> 
> suicide_rule allowed
>                   not_allowed
> 
> komi_rule fixed
>                free
> 
> > What are the major differences?   I don't have a problem with komi being
> > expresses separately.
> >
> >   Here is a little try:   J = Japanses scoring
> >                           C = Chinese
> >
> >                           0 = simple KO
> >                           1 = positional super ko
> >                           2 = situation super ko
> >
> >                           S = suicide allowed
> >                           N = No suicide
> >
> > So you could have  "C1N" as the rules cgos uses.   Strict tromp/talor is
> > C1S", etc.
> 
> Your system reminds me GMP. "One rule, one command" is easier to
> parse. Also, is easier to extend with new rules in future GTP versions.
> And more important is easy to understand to a human, although not
> knowing the command or arguments.
> 
> > I'm sure this doesn't consider all the possibilities, but it at least
> > gets you close.
> >
> > I don't like the idea of having a different nomenclature for each kind
> > of rule-set, such as "Ing", "Australian",  "Tromp/taylor" etc.
> 
> I agree with you now. A ruleset name is even less flexible than your
> system. Also you have to investigate what Ing/Australina... rules are.
> 
> GTP doesn't try to be compact. It tries to be easy to parse for programs
> and easy to read to humans. I think GTP is now in the right way. If
> someone needs a compact protocol should use GMP.
> 
> Andrés Domínguez
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/