Re: [boost] A BOOST GUI Library

From: Bjarne Stroustrup (bs_at_[hidden])
Date: 2001-04-02 15:00:59

Beeman writes:

> 
> The full context of this was that in a panel led by Nico Josuttis, Dietmar 
> Kuehl, and myself, a straw vote was taken.  (Numbers are from memory and 
> very approximate.)
> 
> How many would like to see the Standard Library include:
> 
>      * A low complexity GUI library:     A handful (5?)
>      * A medium complexity GUI library:  A larger handful (8?)
>      * A high complexity GUI library:    A handful (5?)
>      * No, don't include a GUI library:  Lots (>30)
> 
> This is a marked contrast to other possible libs discussed, such as 
> persistence, threads, and so forth.  All of these were greeted 
> positively.  The only one we took a straw poll on was a GUI library, 
> because it was the only one that seemed controversial to the audience, 
> which appeared to include many experienced developers.
> 
> So there is a storm warning flying on this one.  That doesn't mean Boost 
> should steer clear, but it probably does mean rough waters ahead.
> 

Amen.

IMO there are two reasons for the outcome of that note:
        (1) many believe that we can't do it
        (2) some believe that if we did it, it could be an albatros because a 
boost.org library would not be able to match the look and feel of commercial 
systems to be of real use for anyone producing commercial software.
        
(1) may be true, but that would only be a reason not to try if we feared that 
the effort would siphon off effort from more important projects

(2) is a real danger. A language-specific GUI was one of the things that killed 
Smalltalk. You can't build a better GUI than XXX if XXX is the definition of 
correct. Maybe the web and look&feel standards have changed that - I really 
don't know - but this has to be considered. Simply producing a much more 
elegant, more efficient, and easier to use GUI that (say) MS isn't sufficient 
for success. 

        - Bjarne
Bjarne Stroustrup, http://www.research.att.com/~bs