Re: current development

[Philippe Michel]

On Sat, Jan 25, 2020 at 10:15:16AM +0000, Sarah Payne wrote:

> In my opinion, a phone version would broaden appeal / access but the 
> most critical issue is the neural nets. XG feels like a very different 
> animal as an opponent - noticeably more opportunistic & aggressive, so 
> some degree of congruence asap seems critical for gnubg to hold 
> ground.

As other have already pointed, this "feeling" is most probably illusory 
and calling that critical seems excessive.

On the other hand, I think there are some issues that could legitimely 
be called critical. These are situations where GNUbg could end excluded 
or be greatly handicapped for some use cases. 

Two such issues that come to mind are :

- the lack of a maintained packaged Mac OS version. You wrote earlier 
about trying to recruit programmers. What about trying to specifically 
find a Mac hobbyist, or a developer or consultant interested in doing it 
as training ?

- the inability to read XG files. It looks like matches tend 
increasingly to be published in this form and that people often don't 
bother with mat/txt exports.

A few years ago Michael Petch made available a Python library with 
routines to decode XG matches files (maybe not single positions) but he 
didn't go all the way to a small utility that would convert them to a 
sgf file.

Converting it to C and using it in gnubg itself is probably not trivial 
but completing the Python stuff to at least recover the moves, not the 
evaluations, shouldn't be too difficult.