Re: Size negotiation...

[Michael Meeks <michael helixcode com>]

Hi Mathieu,

On 29 May 2000, Mathieu Lacage wrote:
> I believe it quite possible that someone one day implements the bonobo
> interfaces with KDE/QT in which case the size-negociation is -not to say
> the least- _needed_.

	Of course, but it can go via. the X socket/plug. You are still
going to need a socket/plug and I can't see why a toolkit would implement
a socket/plug with no builtin size negotiation. Either way, it is not a
difficult thing to add it seems.

> >     Bonobo's control interface relies on a secondary transport
> > mechanism which is in this case Gtk/X, in the abscence of Gtk/X you will
>
> I do not agree. Bonobo relies on an X window magic number.
> Windows has a similar mechanism.
> I don't know for other systems but I believe that all windowing
> systems have such a similar system of Window identifiers which
> alow you to write from an app to another app windows.
>
> This means that the only place where Bonobo is dependant on GTK/X
> now is the size-negociation and this can be corrected so easily that
> it drives me nuts.

	Well, it doesn't worry me whatsoever at all. If you want to
re-write, re-test etc. code that already works well fine. I would remind
you that you will have to write a BonoboPlug / BonoboSocket that inherit
from their gtk counterparts to make sure we don't have 2 competing size
negotiation mechanisms, since this caused lots, and lots of grief in the
past. But I agree, it would be nice to kill the Gtk/X dependency, it is
just a matter of man hours and serious testing.

> > have to re-write the GUI transport layer, which happens to also propagate
> > sizing information. I see no problem  here whatsoever apart from a couple
>
> The problem here is that the Windowing layer does not propagate the size
> information and that GTK size-negociation mechanism is NOT universal which
> means that if we have such a method which depends completely on GTK, we
> lose a great deal of the interest of the IDLs since implementing them is not
> enough to get a conformant implementation.

	It will never be enough; you need X too.

> Enforcing implementors to use GTK is definitely not a very good idea.

	But we do that already it would seem to me, what is the problem ?
 
> In the meantime, waiting for such a system to be implemented, my feeling
> is that we should implement the current interfaces. I understand your
> reasons but my feeling is that having something relying on GTK in Bonobo
> implementation is just plain evil.

	The Object system ?

> It just drives me nuts to see such a thing: the perfecteness of bonobo
> being spoiled by this...  ;)

	Well, I see your point. Do feel free to write the code, but do not
commit it, post it here and we'll chew it over. It is just a large amount
of work for no percievable gain to fix a drop-off in a transport that
someone is going to have to re-implement elsewhere anyway if they want to
use another toolkit.

	Regards,

		Michael.

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