Re: [boost] Re: A BOOST GUI Library

From: Gregory Seidman (gseidman_at_[hidden])
Date: 2001-04-03 13:46:35

Chuck Messenger sez:
[...]
} As a practical matter, the reason you want a cross-platform GUI
} framework is to write cross-platform apps, which means you need
} all that other cross-platform stuff, too.

Absolutely true.

} It is indeed very messy.  It's not really the sort of pure problem
} which Boost libraries are mostly oriented toward.  It's more
} like a huge collection of hacks -- standardized hacks.

Mm. I once heard Andy van Dam, in his introductory class, claim that all of
graphics is hackery that just happens to work. Regardless of the hackery
underneath, however, the interface can and should be as elegant as the rest
of Boost.

} Starting from scratch seems insane.  The problem space is just
} too big and messy.

I respectfully disagree. Unless "from scratch" means starting from the
framebuffer, what is needed is no more (and no less) than an elegant,
STL-integrated, C++ API wrapper around existing native toolkits. Even
easier would be such a wrapper around an existing cross-platform toolkit
like wxWindows.

} Hopefully, there will be a gelling of support behind one of the
} existing frameworks.  I like wxWindows basically because it
} seems to be quite active.  You get to leverage the efforts of
} lots of other people.
} 
} However, some of the other systems look very interesting -- more
} powerful, but less support.

It is important to separate the boostified interface we might like to
provide from the underlying toolkit it wraps.

[...]
} Maybe a better goal would be to push Boost into existing GUI
} frameworks.  I, for one, would be very attracted to a GUI
} framework which embraced Boost.  I imagine many other programmers
} would be, too.

I would be delighted with Qt if they would abandon backward compatibility
and compatibility with non-compliant compilers; if they were starting from
scratch right now they would undoubtedly integrate with the STL and maybe
even Boost. Since they almost certainly won't, and since even if they do
the licensing will be incompatible with Boost's goals, I propose starting
from Qt's design rather than its codebase and wrapping some toolkit(s).

}      - Chuck
--Greg