KDE adoption of D-Bus

[Olivier Galibert]

On Thu, Feb 16, 2006 at 12:43:32PM -0500, Daniel Macks wrote:
> On Thu, Feb 16, 2006 at 06:24:56PM +0100, Olivier Galibert wrote:
> > On Thu, Feb 16, 2006 at 03:53:57PM +0100, Lennart Poettering wrote:
> > > If this has a realistic chance to be merged, I would be willing to
> > > contribute a patch that uses dlopen() to access the X11 display.
> > 
> > Dbus/session is pretty much designed to be started from the X startup
> > anyway, and to be used essentially in GUI-backed programs.  Even
> > dbus/system and hal are very much for GUI programs too.  So well, a
> > libX11 dependency is not necessarily problematic.  While dlopen code
> > could be, long term, and makes portability way more difficult.
> 
> Actually, a hard libX11 dependency *is* problematic because there are
> other GUIs besides X11. If KDE relies on dbus and dbus has an intimate
> relationship with x11, then it becomes very difficult to port KDE to a
> non-x11 GUI models. If x11 usage goes into dbus (in whatever
> implementation), could it please be made optional at the level of the
> dbus build?

We're talking here about autostarting of dbus/session on possibly-x11
sessions, not usage of libX11 in dbus once it's running, so the
dependency can very much be compile-time optional without loss of the
main functionality.  And if fact an additional point against a dlopen
implementation is that you're hiding the dependency.  How many servers
are you going to start under a native macosx port with X11 installed,
or other cases of systems with compatibility X11 libs installed.

  OG.