Re: [Chicken-users] cross-platform gui toolkit

[Brandon J. Van Every]

Daniel Sadilek wrote:
> > Right, too. Really, "targeting" Qt(4) is a perfectly reasonable
> > choice. Actually,
> > I have no clue what's the best option - but I fear committing to one 
> > particular
> > toolkit.
>
> I see your point. However, independence of the toolkit ranks second
> for me. Having a platform independent API with which you can build
> full-featured professional looking GUIs ranks first.
> It seems to me that here are two groups. One wants a slick API for
> 3D-OpenGL-game-development and the other a comparatively big API with
> lots of standard-widgets for standard-looking desktop-applications.
> Perhaps we should distinguish between the requirements these two
> groups have? I suppose that two different APIs are necessary.

There are at least 3 camps:

- put OpenGL windows on screen, no widgets.  For game developers who 
don't want a bunch of extraneous stuff in their way.  Not much to 
discuss about requirements here.  We've got 3 people working on it, and 
we all know what we're doing well enough.  We welcome anyone who 
actually wants to cut code.

- minimalist GUI, minimal core set of the most useful widgets.  Intended 
to be simple, portable, and sufficient for modest GUI apps.  Not 
intended to be fully featured.  Not important to have native look and 
feel.  Has many different GUI toolkits as backends.  This is Felix's 
plan for conquering the universe.  I fear he won't actually secure the 
support resources for more than 1 platform, in practice.  Multi-platform 
will exist as proof-of-concept, but outside the de facto area of core 
strength, the builds will suck and be forever broken.  I point to 
Chicken itself as evidence, pre-CMake.  MinGW simply didn't get the 
love.  MSVC had 2 different build methods and different path conventions 
than the Unix builds.  It took 1 year to pave all that over.  People who 
come along to do that much gratuitous refactoring are the exception 
rather than the rule, i.e. I am insane.

- maximalist GUI.  Maximal number of widgets, fully featured, can handle 
any enterprise class problem you have in mind.  Native look and feel.  
"Fear of committment" is silly here.  WxWidgets, GTK+, and Qt have been 
dealing with the "heavy duty and portable" problem for the past decade.  
If you want to be fully featured, you can't do better.  Pick 1 and wrap it.

Also there's an undercurrent of research and exoticism in some people's 
posts, with regards to event handling and other would-be GUI features.  
This could skew people into yet more camps.

R&D is a non-issue for OpenGL windowing.  We're not going to do enough 
to get ourselves into trouble.  I'm sure Thu will cook up some event 
handling experiments, and at some point I'll try my own hand at it.  
Don't know what John has in mind here.

For Felix's portaGUI, the R&D impulse will cause a lot of design 
discussion.  Hope it sheds more light than heat.  I don't expect all 
designers to last through the process.  The direction taken will not be 
equally fascinating to all parties, and some will lose interest.

For a maximalist GUI, the R&D impulse is counter to getting the job 
done.  WxWidgets, GTK+, and Qt are exceedingly mature projects with 
gigantic user communities.  "Not Invented Here" isn't going to get 
anyone anywhere.  People have to value the pragmatic result, not the fun 
of doing their own R&D.


Cheers,
Brandon Van Every