Re: GEP 9: merging bonobo-activation & libbonobo

[Michael Meeks <michael ximian com>]

Hi Mark,

On Wed, 2003-03-12 at 20:35, Mark McLoughlin wrote:
> 	I think it would be worthwhile to cc this to another list. There would
> be a lot of people with opinions on this but not on
> gnome-components-list. I've cc-ed desktop-devel-list.

	I believe I spammed every man/woman/child on the GEP list, but then
again I might have reflex discarded it on the admin page; hmm.

> 	You say "two seperate modules creates more problems than it solves".
> The two problems that you give are "packaging is hard" and "I want to
> use BonoboObject. Are there no other problems ?

	Well - there's the problem that before adding any new features to
bonobo-activation-server I want to re-factor the code so it's legible,
and non-cumbersome. I think it's a pretty big problem personally.

> 	I'm not sure either of these or really good reaons for merging the two
> and thus forcing people to use libbonobo where they may just want
> bonobo-activation. But then again, bonobo-activation hasn't seen much
> use without libbonobo.

	Sure - the only person who I thought might object was Bill H, but after
talking to him this appears not to be a big issue. Certainly it won't
affect anyone linking from A to B, and IMHO it doesn't affect any
potential portability, so ...

> 	FWIW, the GObject adaptor would provide a solution to the "I want to
> use BonoboObject part". Granted its not finished, but I'm willing to
> do the last bit of work required ...

	Well; as you know I still have major reservations about the GObject
adaptor. Particularly because it will provide yet more confusion in the
ORBit2/libbonobo space, but also crucially because I don't like the
fairly large non-typesafe extra amount of ORBit_gmethod_register typing
in the C code - instead of just including the IDL generated epv
structure [ something that IMHO BonoboObject gets right ]. Of course,
there are some good points about it - and I'm not altogether oppose to
something like this in ORBit2 - but I'm really not happy with this yet.
I see far more usability mileage in providing extended GSignal/GObject
property proxies in libbonobo personally.

	Finally, I don't particularly want to block bonobo-activation
re-structuring on the possible completion of the a new ORBit2 adaptor,
and I'd far rather that work was put into eg. ORBit2 asynchronous impl's
than this :-) but ... 

	So - that's why I think this is a separate issue, orthogonal to
merging.

	Regards,

		Michael.

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