On Wed, 14.07.10 16:34, Bill Nottingham (notting(a)redhat.com) wrote:
Lennart Poettering (mzerqung(a)0pointer.de) said:
> > The issue is that this is a behavior change (from both sysvinit and upstart)
> > that will need code to be handled properly in other packages. Anaconda,
> > at least, will need to be patched to set the default bootup target
> > differently depending on which init system it's installing, which
> > is kind of ugly.
>
> Hmm, can you please elaborate on what you think anaconda should be
> writing into that file?
ananconda is what sets the default runlevel post install. It does this
by editing inittab. This change would mean that anaconda would conditionally
need to do something else, depending on what it happens to be
installing.
Or it could just do both changes. That should be easy enough.
Hmm, I wasn't aware that Anaconda even asks a question about the
runlevel. Given that I am too lazy to try this out now, what exactly is
this question? i.e. does it ask "Are you installing a server or a
deskop?" or what does it ask?
> sysvinit allows you to have exactly 4 general purpose runlevels,
which
> have the superbly descriptive names "2", "3", "4", and
"5". And that's
> where the story ends. In systemd the model is much simpler, more obvious
> yet more powerful: target units can be labeled freely and you can have
> as many as you want.
It allows 7, 8, 9, as well. (Wheee, undocumented features.)
> You really cannot see why runlevels are old cruft and completely
> arbitrary in their design and naming, and really need replacing?
Oh, they are. But they're also 30+ years of historical precedent that
people are used to and code is expecting. Even OSX has them in some
manner.
And that's why we support them to a large degree, via compatiblity
targets "runlevel3.target" and similar. What we don't support is setting
their default choice with /etc/inittab. That's all.
Lennart
--
Lennart Poettering - Red Hat, Inc.