Re: [boost] [rfc] Gui library

From: Jurko Gospodnetic' (jurko.gospodnetic_at_[hidden])
Date: 2007-10-21 10:02:09

   Hi Felipe.

> I'm already working on a GUI library for some time (three weeks). And
> come up with some designs and code.

   We use wxWidgets for a portable GUI framework and it works fine for 
us. Good points are:
   - is open source
   - has no license problems that we ran into
   - has a huge user base
   - is extremely active
   - has been extremely active for a long time now
   - releases new versions reasonably often (unlike boost :-))) *gdr*)
   - the documentation is good enough (as an extremely good introduction 
for regular usage there is even a paper book published)
   - written in a good enough style that it has been easy for us to fix 
and supply patches for a few problematic areas we ran into.

   On the 'bad side', wxWidgets does not implement just a GUI library 
but also has its own 'conquer the world' mentality so it has parts like 
its own threading, network or file system support but we do not use any 
of those areas.

   I do not mean to infer that no one should explore implementing the 
'ultimate GUI library' but from my viewpoint it seems more reasonable to 
hook up to this project and mold it into something better if something 
does not suit you. The whole GUI support has so many niches and platform 
specific implementation details in it that I do not believe one can make 
it ideal but only 'good enough' and strive to constantly improve it. And 
  an existing library such as wxWidgets already has a giant lead in the 
right direction.

   Starting a boost.GUI library just for the sake of 'implementing the 
whole world as a part of boost' does not seem like a good way to go. One 
possibly good way I see to go about this would be to merge with an 
external library such as wxWidgets, the same as was done for the 
(excellent!) asio library, with great benefits for both sides. Boost 
would get GUI support and the GUI library would get the benefits of 
existing excellent boost libraries and not have to reimplement them 
(thus removing the need for double maintenance efforts). Also, users 
would find it easier to learn/access both of the libraries.

   Just my 2 cents worth...

   Best regards,
     Jurko Gospodnetic'