F21: why Fedora still has not alternative init?

[Frantisek Hanzlik]

Gordon Messmer wrote:
> On 05/03/2015 05:04 AM, Frantisek Hanzlik wrote:
>> - (bigger harm) Why hasn't Fedora alternative (upstart/openrc) init?
>> ...
>> When systemd presents itself as compatible with sysvinit, then IMO
>> having alternative init in Fedora should not be too big problem.
> 
> Systemd is backward-compatible with SysV init scripts, but other init
> systems are not forward-compatible with systemd unit files.  If Fedora
> were to support an alternate init system, it would have to ship both SysV
> init scripts and unit files with all of its daemons. Then the developers
> would have to sort out how to be both a) backward compatible with SysV and
> b) ignore the SysV init scripts in favor of unit files when systemd is in
> use.  Bugs in daemons might show up under only its unit file or only its
> init script, which would increase the complexity of handling bug reports. 
> So, complexity is one of the reasons that there's not an alternative init
> system.

SysV init scripts are here for ages, IMO is not problem dust up and
revamp them. In long-term distros as RHEL 5/6 they will be with us for
years...

> Another one is that systemd enables a handful of Linux features that other
> init systems don't (e.g. cgroups).  Any package that relies on the use of
> those features might be broken on another init system. Or it might simply
> behave in a way other than the documentation suggests, which would lead to
> bug reports that are associated with the lesser init systems.

We have lot of alternatives in Linux system (several desktop WMs etc)
already, alternative init is in this case (IMO) just small piece of all
system SW.

Although I understand that there are features which are interesting for
some Linux users, they are not too important to me. I was completely
satisfied with SysV init/upstart in previous Fedora releases, they works
without problems for me already.
What I sometimes needed was to define several runlevels and in them to
run various daemons - and it was much easier to do using SysV init
runlevels.

>> - (smaller harm) Why hasn't systemd option to run without journald?
>> ...Why then in my system must run journald daemon, quite useless,
>> occupying 2.5+ MB of memory?
> 
> I tend to think that's a better question.  2.5M of memory is trivial, but
> I have systems where the RSS of systemd-journald is 30M+  The very high
> variability of the memory size for that process makes me worry about
> memory leaks.

Not sure when 2.5MB is little or plenty (several years ago I work as
programmer at PDP-11/RSX-11M systems, which had ~ 1 MB memory - and
there was running _lot_ of user sessions simultaneously, which control
the factory production). Despite of that, why I should run (in my case
useless) journald daemon, when I do not want it? And must use complex
settings as
systemd.log_target=syslog-or-kmsg/LogTarget=syslog-or-kmsg/DefaultStandardOutput=syslog+console/Storage=none
etc, when should be
sufficient simply disable journald?
-- 
Franta Hanzlik