[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: On init in Debian

Stéphane Glondu <glondu@debian.org> writes:
> Le 30/03/2012 08:18, Tollef Fog Heen a écrit :

>> I doubt you'll get upstreams to write metainit files.  I think we'll
>> have upstreams providing systemd files and so I think metainit will
>> basically be #15 in http://xkcd.com/927/.

> Actually, it's more systemd that looks like #15.

I don't think the analogy really works on either count, although it's
somewhat closer for MetaInit.  But it definitely doesn't make sense for
systemd.  systemd's goal wasn't to become a standard that supported things
people were already doing.  Rather, both it and upstart were aiming to
incorporate into the init system brand-new functionality that wasn't
currently supported at all, things like real process monitoring beyond
init's meager capabilities, safe process kills without using PID files,
and of course event-driven boot.  They're effectively two different
strategies and projects aimed at solving the same set of technical

The difference between this and standards is that standards as commented
on by XKCD are mostly looking to collect existing solutions to problems
that are currently solved in mutually-incompatible ways into a uniform
framework.  They usually aren't intentionally breaking new ground.

The maintenance of systemd is actually quite the opposite of a standard.
It's focused on being clean, supportable, and fully integrated with Linux
capabilities, *not* to solving everyone's use case, even to the detriment
of being universal.

A standard in the init script space would look more like what the LSB says
about init scripts: a conservative standardization of well-known and
pre-existing ideas and technology that have been indepedently solved
already by multiple different systems in ways that aren't interoperable.

Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>

Reply to: