Re: using dbus in the platform

[Havoc Pennington <hp redhat com>]

Hi,

The point about toolkit vs. app framework I think is perfectly 
debatable, but is best debated on its own, and considering lots of 
things besides dbus, imo. I won't try to start that thread by laying out 
a comprehensive position statement or anything. There is already some 
stuff on "Project Ridley" out there, and already GTK has examples (such 
as the Beagle/Tracker integration), as a starting point.

Matthew Johnson wrote:
> I also agree that D-Bus should be standardised as a protocol rather than
> an implementation. I believe it's the only way to make it a universal
> standard, rather than something some people happen to use.

Is this relevant to the dbus-in-gtk dicussion, however?

As you say "D-Bus should be standardised as a protocol rather than
an implementation" and that is in fact the case and intent, modulo work 
that simply has not been done but anyone is encouraged to do (i.e. the 
spec could be better).

However, if we say "X11 should be standardized as a protocol rather than 
an implementation," I don't think it follows that GTK should not use an 
implementation of X11. Clearly GTK has to use Xlib, XCB, or at least 
some internal equivalent. It could use a pluggable implementation, 
perhaps, but that would still be an implementation, and the plugin API 
would place constraints on viable backend implementations.

In my opinion the dbus (protocol) situation is exactly the same as the 
X11 (protocol) situation for GTK. The only difference is that X11 is 
used by virtually all - but not all, i.e. DirectFB - Linux-based 
environments, while dbus is used only by GNOME, KDE and XFCE and perhaps 
not some other setups. However, neither X11 nor DBus (protocol) is used 
by GTK on all its targets (Windows, Mac, obviously).

Havoc