RE: [computer-go] GTP and Tourney in SmartGo 1.4

[A van Kessel <adriaan.van.kessel@xxxxxxx>]

Mark Boon wrote:
>  My question is though: are you 100% sure that any future addition to GTP
>  doesn't break any or all the parsers already out there? (Which IMO are all

That has to to with the semantics of the protocol.
With XML, syntax and semantics are separated: the syntax is
fixed, and the semantics are based on some definition, which can
can be changed anytime.

GTP is line-oriented, Baically modelled after SMTP: a request
is a line, a response is one-or-more line, plus an extra empty line.

In both cases the "message" can contain anything, sending HTML
to your proposed XML interface will (hopefully) do just nothing,
sending "EHLO" to the GTP- "parser" will probably just answer
an "? unknown command\n\n" .

For GMP, things are basically the same: an unknown (must be extended)
command can only be ignored.

In any case, if a command is added to the protocol (or the
protocol is changed in another way) , any program using the
protocol has to be changed in some way: the appropriate
functions have to be called (created?).

Stability (over time) is more important to a protocol
than extensibility, IMO. There a plenty of ways to extend
a 80 character line.

I see no (real) problem with GMP, and no (real) need for XML.
As david Fotland said, there is no need to combine your
Game-Playing-Protocol with your method of regression-testing,
or otherwise communicating with your program.
You may create a XML interface, but don't force me
to create one ...

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