Bug#727708: Call for votes on init system resolution
On Thu, Feb 06, 2014 at 07:42:41AM +0100, Lucas Nussbaum wrote:
> On 05/02/14 at 22:41 +0000, Thorsten Glaser wrote:
> > I think it is not up to the d-i people to decide on the init system
> > anyway – especially as not d-i but debootstrap is the canonical way
> > to install Debian… and debootstrap goes by whatever ftp-masters put
> > into the override files, and whatever package dependencies and meta
> > information (such as Essential: yes) there are.
The latter's true, yes, but this is a distinction without a difference.
debootstrap has been maintained by the d-i team since 1.0.0 in 2007.
> > So, “jurisdiction overlaps” seems to fit quite well.
> As far as I remember, I don't think that any of the d-i maintainers,
> debootstrap maintainers, ftpmasters, or sysvinit maintainers have
> claimed that it was their sole jurisdiction to decide on the default
> init system. So I agree that there's jurisdiction overlap.
Right. A simple thought experiment to get a first approximation of
jurisdiction in Debian is to look at who might be able to put a given
change into effect as part of their ordinary work, changing only the
things they're generally agreed to "own". Using that we get at least
* ftpmaster could change Priority fields to cause a different init
system to be the default
* the debootstrap maintainers / d-i team could decide to act at
variance with the Priority fields and install a different init
system; there's plenty of precedent for not going just by Priority,
and we might reasonably want it to have options to do so in this case
* the sysvinit maintainers could decide to throw in the towel and make
sysvinit a transitional package for some other init system
[hey, I didn't say all these options were realistic]
* any boot loader maintainer could decide to tweak their default
configuration to pass init=something (indeed for a while I thought
that that might well end up being the implementation mechanism for
the result of this vote, although I've since been cluebatted
* the maintainers of a sufficiently widely-used package could cause it
to depend on a given init system
So, yes. I would be interested in whether the Secretary agrees whether
this is jurisdictional overlap. If the answer is no, then it would be
very helpful to have examples of what would be so that we can act
accordingly in the future.
Colin Watson [email@example.com]