CNI for C

Wed Apr 7 19:56:00 GMT 2004

>>>>> "Geert" == Geert Bevin <gbevin@uwyn.com> writes:

Andrew> * We can't use VM-specific features that may be useful, such as CNI.
Andrew> <http://ometer.com/desktop-language.html>
Geert> I can't speak for others, but I hear all the time that Mono is
Geert> not about being compatible with .Net and simply to provide a
Geert> nicer environment for Gnome development. If you continue on
Geert> this line of thinking, you can say the same for GCJ.
I hear conflicting stories about this stuff. It would be nice to have
more clarity from the Gnome community about what they really want...
For instance, if compatibility with Sun's Java doesn't matter, then by
all means let us implement what is desirable. This might mean writing
some simpler way to make native bindings a la PInvoke (we already have
all the needed low-level technology, it is really just a matter of
building some bits into the compiler and runtime). This could even
extend as far as adding features to the language that are very
desirable (the GUI hackers seem to like delegates quite a bit, we
could add those in some backward-compatible way).
On the other hand, Sun is a stakeholder in Gnome. Perhaps they will
reject a modified Java. But wouldn't they reject .NET on the same
basis?
Geert> What I do know is that the easy C integration is making Mono and C#
Geert> gaining more and more momentum, while a perfectly functional
Geert> free-from-patents technology exist with GCJ. As I see it, easy
Geert> integration with C is a minor addition to what is already a very stable
Geert> and comprehensive platform.
Yes, I think we've been quite bad at pushing our technology. Are gcj
and Java perfect? No, of course not. But I think they are attractive
along many axes: there is a lot of free Java code, including
development environments and other "big" things; gcj is quite mature,
has good performance, and is not only extremely portable but in fact
very widely ported.
So, how can we get the word out more effectively? I don't know. I
think we need some evangelists in the Gnome world (and other
communities) to help out. And I think we need to listen to those
folks when they talk about what it is they need.
Considerations like these are why I think java-gnome is such an
important project. Other bindings would also be useful, to make it
easier to write Gnome-ish applications.
Another complaint I hear sometimes is that our release cycle is too
slow, since we're tied to gcc. We could probably start doing more
frequent libgcj releases if that would help.
Tom