D/Bus / CORBA moving forward ...

[Michael Meeks <michael ximian com>]

Hi there,

	As you may or may not know; there has been a particularly
unhelpful and unproductive slanging match inside Gnome recently
wrt. the future of component technology. I must confess to having been
particularly bitter at having been deliberately not involved in the
D/BUS discussion, despite having struggled to maintain and improve the
Gnome infrastructure in this area for years. However, it's past time
to for me to put that aside, and move forward. Apologies for my tone -
it's hard to see many man months of your work writhe defenselessly on
someone else's chopping block.

	There is a lot about D/BUS that appears good - particularly,
having a system wide messaging bus. It's excellent that it (seemingly)
has the support of both KDE and GNOME communities. It seems likely to
become an important part of Linux systems.

	On the other hand - there is a largish amount of existing code
in Gnome, now, that needs maintaining. There is also the scope issue -
Bonobo particularly aims to tackle a slightly larger problem space
than does D/BUS. It is also certain that D/BUS has a long way to go
before it reaches the level of hard-core API/ABI stability, and
feature completeness that all users must demand of it to be in the
platform at the level at which it is targeted.

	At times like this there is bound to be an amount of
uncertainty and doubt about the existing, working but non-trendy
projects. I personally will continue to maintain and improving the
CORBA / bonobo infrastructure, and attempt to encourage people to use
/ adopt it - it will be with us essentially forever as part of the
Gnome API/ABI compatibility guarantees, and it's great to see the
plugging power of components inside nautilus, and hopefully soon
evolution-1.4.

	I also intend to address some of missing issues inside Gnome,
as detailed by Havoc in his seminal D/BUS post:

https://listman.redhat.com/pipermail/message-bus-list/2002-November/000004.html

	Where new functionality is needed, I propose that a common
sub-set of it be wrapped in such a way that the underlying technology
(bonobo/CORBA) is concealed - but there is much that can be done to
make components work. I also plan to consolidate bonobo-activation and
bonobo to allow sensible code re-use.

	To show good faith, I hope to engage with the D/BUS
discussion, review the code / direction, and provide whatever insight
I have into ways to help it succeed. It is my conviction that creating
a robust, friendly IPC system with which people can interact is a more
complex and intricate thing than may appear at first sight. In the
end, I think you end up having to create a type system, and an IDL
like contract definition language. A brain-dump to follow at some
stage.

	So: in summary; this is not a binary choice - it is possible
for both systems to exist, and work together to give users a better
desktop experience.

	I look forward to working more pleasantly with many of you,

	Regards,

		Michael.

-- 
 mmeeks@gnu.org  <><, Pseudo Engineer, itinerant idiot