Re: What is GNOME office?

[Havoc Pennington <hp redhat com>]

Hi,

I agree with Dom; I think Abi is plenty GNOME-aware. I'm actually not
sure the race to use all the unstable GNOME libs is such a great idea;
as I said I think we should be thinking a couple years at least for
finishing the suite, even if we get tons of full-time developers on
it, so probably we'll have e.g. substantial changes to the component
model by then, better support for .NET-style shared runtime, X render
extension with a virtual render API in GTK, accompanying better fonts,
etc. Abi does have a nice abstraction barrier between their stuff and
GNOME technology flux.

The AbiWord approach to cross-platform doesn't require people hacking
on platform A to understand platform B, which is a good thing.  The
main disadvantage to their approach is that the platforms don't stay
in sync automatically, but free software development is well-suited to
doing the typing to keep them synced. The other disadvantage I would
say is that it does add a higher barrier to entry in working on the
code, since you have to learn the Abi framework in addition to GTK.
But given the strong AbiWord community this doesn't concern me too
much.

If I were doing a suite from scratch, my inclination would be to just
write to GTK, and put one or two people on bringing the win32 port up
to speed and making its emulation of native widgets better, a la Qt. I
think the "least common denominator" toolkit (exemplified by AWT and
Netscape 4) is widely considered to be a bomb, and Swing, Mozilla, and
Qt exemplify the more successful approach of just porting the drawing
layer of the toolkit and using non-native widgets themed to emulate
the native ones. The usual objection to this is that widgets don't
look native, but Qt shows how to fix that. So I'd find the OpenOffice
plan to use a VCL layer over native widgets worrisome, much more so
than the AbiWord approach.  Too many other projects have found
AWT-style toolkits to be junk and had to start over.

Havoc