Switch to systemd?

[Tomasz Pala]

On Tue, Nov 29, 2011 at 13:07:35 +0100, Jacek Konieczny wrote:

> The question is, should we now start maintaining the third init
> subsystem? Or should we drop anything done for Upstart? Or even drop
> legacy SysVinit support?

Legacy SysV can't be dropped, as upstart won't be finished to the level
when it could replace SysV, and systemd requires pretty new kernel.

> Anybody relies on Upstart or can we assume it was a dead end road and we
> can give it up? Do we need any transition facilities (I will take care
> of my systems myself)? 

As upstart was never oficially supported, I don't see a need to make
transition scripts or any tools. But I don't see a reason to drop it as
well! We can't tell today what is going to be the final init. History of
open source is full of dead ends and come backs (recently we dropped
util-linux-NG, pwdutils obsoleting shadow is dead, there was libungif vs
giflib issue, multiple Motif implementations, Image*Magick etc. - oh,
and rpm of course).

> What about current SysVinit scripts. Should we keep them and maintain
> them or start switching everything (including whole rc-scripts) to
> systemd? This can break backward compatibility a lot, but we could gain
> much more consistent (which also means: more reliable and faster) system
> in the end.

IMHO we can't break SysV (and LSB). If rc-scripts can't be expanded
(they should, as most of functions or rc.sysinit is actually pointless for
systemd), it should be forked.

> Anyway, LSB scripts in /etc/init.d should still be supported in some way
> and probably wrapper scripts for services maintained by systemd should
> be placed there too ??? some things still rely on that and this will allow

LSB and SysV are both supported by systemd as long as COMPAT_MODE is
compiled in.

-- 
Tomasz Pala <gotar at pld-linux.org>