Re: Debian Buster release to partially drop non-systemd support
- Date: Tue, 16 Oct 2018 16:08:57 +0200
- From: Adam Borowski <kilobyte@xxxxxxxxxx>
- Subject: Re: Debian Buster release to partially drop non-systemd support
On Tue, Oct 16, 2018 at 07:47:36PM +0800, Benda Xu wrote:
> Thorsten Glaser <t.glaser@xxxxxxxxx> writes:
> > … this applies to the shim only. I was a bit surprised seeing on
> > someone else’s system that it was no longer installable, but almost
> > all systemd-free systems of people I know do not use the shim anyway,
> > so I’d take the Subject line with a few grains of salt.
> > With only two modified binary packages (policykit-1 and udisks2) I’ve
> > got a complete KDE environment running, except for network-manager,
> > without the shim. People who do not use the full desktop environments
> > (but the leaner ones, or just a window manager, or no X11 at all) have
> > no problems with the current situation.
1. You need to recompile these packages, this is not something an average
user or even sysadmin knows how to do.
2. You lose basic functionality like being able to shutdown from a GUI.
> I was about to reply to this thread, but you have completely expressed
> what I want to say:
> 1. systemd-shim is not necessary, even for DEs (except GNOME3).
Alas, dependency chains say otherwise.
What almost all of those packages want is logind functionality, but I quite
don't get "why?". The only answer I received so far is "multiseat".
Except, logind doesn't allow multiple GUI logins on the same machine for the
same user no matter how many seats you connect from (this worked correctly
before!), and it offers additional features for non-GUI logins either. So
all this functionality does is telling you that user X is logged on via
vt/pty Y. I recall a command named "who" on stone age Unices, and wonder
what the whole hassle is about...
But, it's not entirely unreasonable to provide the dbus interface logind
users want. I'd prefer such an interface be a shim to standard Unix (like
who -- utmp), which would have the added benefit of working on BSD and Hurd
(which I don't personally care about but some do) -- yet such an daemon
would have to be written. I for one don't volunteer to write it by
tomorrow... Thus, let's use elogind which actually exists.
Other pieces are some power management that has been moved into logind, etc.
> 2. sysvinit-core is very stable and do not need new uploads.
Exactly! Almost any changes in recent years were entirely because "systemd
wants X". Thus, no wonders people are angry because of unnecessary work.
⣾⠁⢰⠒⠀⣿⡁ 10 people enter a bar: 1 who understands binary,
⢿⡄⠘⠷⠚⠋⠀ 1 who doesn't, D who prefer to write it as hex,
⠈⠳⣄⠀⠀⠀⠀ and 1 who narrowly avoided an off-by-one error.