Re: initscripts: using "/etc/init.d/<package>"
On Sun, Jul 06, 2008 at 05:36:10PM -0700, Russ Allbery wrote:
> Steve Langasek <email@example.com> writes:
> > On Sun, Jul 06, 2008 at 04:38:26PM -0700, Russ Allbery wrote:
> > sysvinit is:
> > Package: sysvinit
> > Essential: yes
> > Pre-Depends: [...] sysv-rc (>= 2.86.ds1-1.2) | file-rc (>> 0.7.0), [...]
> > And both sysv-rc and file-rc provide invoke-rc.d. Since sysv-rc was split
> > out of the sysvinit specifically to allow the alternative with file-rc, I
> > think it's still intended that invoke-rc.d be regarded as an Essential
> > interface?
> Hm, yeah, probably so.
> Should we remove the fallback code in the Policy example and recommend
> people use invoke-rc.d unconditionally?
I think that would be better.
It makes me think, though, that we don't really have a place to centrally
document what interfaces we consider essential and what ones we don't.
Effectively, anything provided by a package that's a pre-dependency of an
essential package can be treated as essential by this same logic, with
sometimes unintended consequences if a maintainer later wants to drop an
interface. So perhaps we should be explicitly blessing invoke-rc.d as
virtually-essential, with the consent of the sysvinit maintainers, before
changing the example?
(For that matter, /usr/bin/awk is another "virtually-essential" interface in
this category, and we've held forever that you don't have to depend on awk
because base-files does this for us; but that's also not documented in
policy, that I can see.)
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
Ubuntu Developer http://www.debian.org/