Date: Tue, 6 May 1997 00:01:59 +0200 (MET DST)
From: Arno Hollosi <hollosi@sbox.tu-graz.ac.at>
Subject: Re: SGF


On Mon, 5 May 1997, Jens Yllman wrote:

>  I was just looking at the 'EBNF definition' of SGF. And there it says
> 'GameTree = "(" Sequence { GameTree } ")"'. Does this mean that there
> can not be any nodes between and after the GameTree? I know I sound

Correct. No nodes in between gametrees.

> avoid the kind of trouble we've had with diffrent implimentations of SGF
> before. And then we should be very clear of these things. So everybody
> writes files everybody can read.

The EBNF definition is clear - there are no ambiguities.
Every programmer has seen EBNF defintions at least once.
I don't plan to extend the description as everyone not able to understand 
this widely used syntax shouldn't be allowed to program anyway.

>  I just have a small question to everybody. If the file you are about to
> save does not contain anything that is new in SGF FF[4], should we save
> it in a older version. I don't think so now, because I think we should

FF[4] can be written in a FF[3] compatible way (and vice versa).
Use 'tt' pass moves, no compressed point lists and no soft linebreaks.
And of course no new properties (e.g arrow or line).
It doesn't matter then, if you write 3 or 4 into the FF[] field.

>  Arno, does your program only convert from FF[1-3] to FF[4] or can it
> check FF[4] files for errors? I only wonder if we can use it to verify
> our programs.

SGFC is intended to be the SGF reference implementation.

SGFC knows all FF[4] properties and all FF[123] properties which were
defined in Anders' theses (FF[1]) or Martin's web pages (FF[3]).

SGFC checks almost all error conditions that can occur in a SGF file.
Sometimes it even warns about bad style.

SGFC is able to check/correct and convert FF[1]-FF[4] files.

And yes, it may be used to very the output of your program.

/Arno