Re: [gdm-list] Branch update

["William Jon McCann" <mccann jhu edu>]

Hi Oswald,

Thanks for your response.

On 10/4/07, Oswald Buddenhagen <ossi kde org> wrote:
> On Wed, Oct 03, 2007 at 08:38:53AM -0500, Brian Cameron wrote:
> >>   3. A freedesktop.org project
> > If other communities (e.g. KDE) are enthusiastic about working
> > together to create a single display manager to rule them all,
> >
> to a certain degree.
> there is some desire to get rid of the (by now remotely) xdm-based
> backend. it isn't that strong, though.
>
> an actual common project would be faced with certain management
> difficulties - the backend + a desktop's frontend(s) need to be released
> with the desktop itself (at least for kde) - i don't think kdm having
> "extragear-character" would be that welcomed.
>
> dependency-wise it is clear that the backend must not use more than glib
> (and gobject, if it must be) from the gnome side (yeah, yeah, we don't
> care how *you* see glib ;). libintl, xlib (and xcb), some minor libs -
> that's it.
> backend configuration must be somehow obtained from the frontend (be it
> an object from the frontend side being linked into the backend). kdm
> does this for ages, btw.

We've split the daemon completely from all the user interfaces.  The
greeter will have a well defined interface (using D-Bus) to
communicate with the daemon.  The idea is to have different greeters
for different scenarios (different distros / desktops).  The daemon
will determine which greeter to run using .desktop files and TryExec.

The dependencies for the daemon are only D-Bus, X11, glib, and gobject.

We can conditionally build different greeters in the same tree.  At
some point we might want to support out-of-tree greeters but we'd need
to be sure we have a stable API first.

> some random thoughts:
> - after reading the status updates i found that you repeated quite some
>   of my thoughts (most of which i never put into practice :}). :-)
>   you may want to read the kdm todo, too:
>   http://websvn.kde.org/trunk/KDE/kdebase/workspace/kdm/TODO?view=markup
> - i always thought gdm's internal ipc over the fully-fledged socket
>   protocol is completely overengineered. d-bus is taking this to another
>   level ...
>   communication with the outside is another thing - i've been
>   considering d-bus since its early days. so far, it wasn't that
>   pressing, though.
> - oh, well, we can discuss the technical details another time.

Sounds good.  If we don't decide to use the same backend daemon code
then at least we should try to standardize the external D-Bus
interfaces.  However, we may find that in making changes to support
D-Bus in that way we may as well share a daemon too - and make a lot
less work for ourselves.  :)

> two things should be clear to you:
> - i accept no regressions from current kdm
> - the bulk of the work would be with you, at least for now

I accept those conditions.  I appreciate you being forthright.  It
sounds like we are all interested enough that we should at least give
it a try - it can't hurt.  The redesign we've discussed here is at
least different enough from both KDM and GDM to warrant a new project.

Unless there are serious objections, I'll go ahead and create a new
freedesktop.org project for DisplayManager.  I'd love to be able to
work together on this.

Thanks,
Jon