Fw: different init-systems

[jean.bruenn at ip-minds.de]

>>> Do we want to provide support for more than one init-system (which
>>> might result in a mess to maintain) like having all 4 init-systems in
>>> moonbase?
>>
>> I'm neither for or against maintaining 4 init systems, just wondering
>> why we
>> should; aside from user choice type reasons.
>>
>> I do think if we had a larger developer pool it would be a reasonable
>> thing to
>> do for the user. Personally I don't think I would want to dicker with
all
>> four.

i kinda agree. Currently, due to the lack of active developers i'd be
against maintaining 4 (or more) init systems. I'm just curious whether
there is a reason to keep sysvinit in moonbase (like, for people who run
very old lunar installations/kernels) or whether we can assume that
switching directly to systemd is a good choice.

>>> Should we compare the different init systems (i'd do if its
>>> worth the work, if everyone says "no we dont want to change" no need
to
>>> do such work)?
>  
> choices currently are limited:
> 
> - sysvinit is as is, it's not going anywhere
> - systemd is the only actively maintained and developed replacement
> 
> the alternatives:
> 
> - upstart
> - init-ng
> - runit
> 
> all seem to be 'niche' replacements that do not come close to what 
> systemd brings to the table as far as really changing the way things 
> work, stability, prospect, documentation etc.

I started to compare the init-replacements (not finished yet:
http://jeanbruenn.info/new/2011/04/05/lunar-linux-init-systems/) and i've
been a big fan of runit. However, right after looking at the information of
systemd i agree, i also think systemd is the way to go. I mean, we have
sysvinit and runit on one side (i dont know anything about init-ng) where
runit is a good replacement for sysvinit (it supports parallelized startup,
is sysvcompatible, has service-dependencies and service-supervision - They
won't add more features if i understood correctly, because runit has
reached a super-stable-behaviour and they want to keep it simple and
stable) on the contrary there's nothing like event-based or
ondemand-starting of services. So we're again at upstart and systemd. At
upstart we first have to face that it's using python, and as we're source
based that'd mean: as soon as python is broken, the init system is also.
Whats left? systemd.

Jean