Re: Proposal for systemd interface reliance

[Lennart Poettering <lennart poettering net>]

On Thu, 26.01.12 10:41, Olav Vitters (olav vitters nl) wrote:

> 
> The other announcement. Related to systemd, but not the same.
> 
> I am right that hostnamed, localed and timedated do not rely on systemd
> to be running right?

That's actually something that my chart should have cleared up. The
rightmost column lists which interfaces can be separated from the core
of systemd. Also see the bit in the "Explanations" section about that.

>  - systemd only systems: ensure these daemons are packaged and running
>  - systemd+sysvinit systems: put these daemons in a separate package and
>    ensure they are running under both init systems. They do not rely on
>    a running systemd.
>  - other init systems: Please write a _new_ daemon to provide this
>    functionality. The D-Bus API is really simple and instead of patching
>    hostnamed/locald/timedated to support other distributions/init
>    systemd, the recommendation is to just write a new daemon which
>    implements the D-Bus API.

Well, even distros which currently do not carry systemd at all
(i.e. Ubuntu) can just compile systemd like normal but only package the
hostnamed binaries (and the others) and not include anything else. (see
the "explanations" section in that chart wiki text about that.)

I mean, there's really no point in being that allergic to systemd that
they couldn't even make use of these mechanisms... It's free software
after all.

Thanks for putting this together!

Lennart

-- 
Lennart Poettering - Red Hat, Inc.