Re: Requiring systemd for the gnome-settings-daemon power plugin

[Bastien Nocera <hadess hadess net>]

On Fri, 2012-10-19 at 17:48 +0200, Antoine Jacoutot wrote:
> On Fri, Oct 19, 2012 at 07:43:26AM -0400, Matthias Clasen wrote:
> > On Fri, Oct 19, 2012 at 5:14 AM, Antoine Jacoutot <ajacoutot gnome org> wrote:
> > 
> > >
> > > I think at one point the GNOME project will need to step up and explicitely states that GNOME is a Linux-only Desktop.
> > > I am a BSD user; don't get me wrong, if GNOME goes Linux-only then so be it. But the current situation is hard for us because it is unclear where all of this is going.
> > > When systemd was first mentionned on the lists, it was said it wouldn't be a hard requirement. Fair enough, we are "only" talking about the power plugin here but the way it is going systemd will soon be needed for more important features.
> > > I'm just wondering if it is still worth trying to maintain GNOME for !linux platforms (like I do on OpenBSD). Implementing some of what systemd provides is far from trivial for us.
> > >
> > > To summarize, it'd be nice to know whether there is still a chance to see GNOME running on BSD in a near future. If everything is going systemd, then the answer is clear, but for now I lack the informationit
> > 
> > Hey Antoine,
> > 
> > I think there's a good chance for GNOME running on BSD, thanks to
> > people like you who keep things working. I can imagine it feels like a
> > sysiphus job at times - I hope you get thank-you letters from
> > BSD/GNOME users every now and then...
> 
> Actually I do, yes :)
> 
> > Bastien is speaking as the gnome-settings-daemon maintainer, and I can
> > understand why he wants to get rid of the complicated maze of
> > talk-to-upower-or-to-consolekit-or-to-systemd. It is his decision to
> > make in the end, but there is certainly enough time between now and
> > 3.8 to evaluate the best way to keep things working on BSD, no need to
> > throw in the towel now.
> 
> Sure, but my initial concern is that once you have one foot in
> systemd, why not embrace it totally?
> If we are talking about implementing a couple of systemd interfaces,
> fine.
> If the end goal is to need most of systemd to have a working Desktop
> environment then I am very much concerned and would love to know about
> it.
> 
> Note that my concern is very selfish I agree, I am using GNOME not
> just as a personnal Desktop but also maintain several thousands
> installations. That's why I am even more interested about the
> direction it is going.
> 
> The way I see it is that people were depending on somewhat proven
> portable technologies (for the most part) and the arrival of systemd
> now splits the community. I don't see systemd as "just another
> dependency", it's deeper than that because it aggregates lots of
> things that could originally be into separate projects.
> Don't see this as a rant against systemd, it's not; I'm just confused
> a Desktop environment has to depend on such specific low-level
> software.

Because we want the desktop to not feel tacked on. Having had to correct
kernel bugs, udev bugs, X.org bugs, etc. to get some GNOME features
working, or fix user bugs, I don't consider "root" daemons to be
low-level. I also don't think suspend or multi-seat/fast-user switching
support is something that you want to tack on. You need integration in
the desktop to do it well.

> If I want to explain it in a very stupid way: why does an
> init/cron/syslog/... replacement is needed to change a timezone or
> track user sessions?

It's not an init replacement, as is pretty clear from the functionality
it provides. Not my battle though.

>  It's not, probably. But the problem is that systemd implements lots
> of these things, it's not the fault of the GNOME project of course,
> but if some of the interfaces were actually separated from systemd, it
> would make it way easier for distributions or BSD systems that don't
> use systemd to implement them and submit portability patches (which
> are not accepted in systemd itself anyway).

Then they would have the same problem that GNOME has now, namely that
they would be held back by portability rather than pushing the
functionality forward.

>  Since this is not the case, I am a bit disappointed that GNOME relies
> on such interfaces...

We rely on the D-Bus interface. You can reimplement that interface on
your system.

> Hopefully my mail will not be seen as a dumb rant, I just wanted to
> express and explain some of the frustration I have experienced ;-)