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

Re: On init in Debian

On Fri, Mar 16, 2012 at 10:14:53AM +0000, Lars Wirzenius wrote:
> There is a lot of feelings and temper involved in the current discussion
> of init implementations in Debian. I'd like to try to de-escalate by
> summarizing things in as objective and non-confrontational manner as I can.

Lars, thanks for this summary. Despite the fact you seem to be unhappy
about having posted it, I find it quite useful. For one thing, it's a
good mail to restart from when thinking back about this whole debate
after a while.

I also have the impression that in discussions that are naturally prone
to "religious arguments", it's unreasonable to expect that *all*
contributions will be constructive (for some personal definition of
"constructive"). There will always be contributions that we don't think
are constructive, or that are simply partisan. That does not necessarily
mean that the whole discussion has been useless. In all these "upstart
vs systemd" threads, it seems to me that we've fixed quite some points.

An important one is the lack of hands on experience with either of the
contender (systemd, upstart) _in Debian_. A number of post in the
threads seem to be just relying arguments from the respective marketing
camps, which naturally attract the ire of people who have actually
worked on the systems and feel the need to debunk myths. A number of
other posts are on rather general principles (e.g. "we should not adopt
something that is not ported on $system"). That is all fine and well.
But we should all know how we like doing things in Debian: we will not
bet the choice of the default init system on something we haven't
tested. Therefore, a more productive use of the time of -devel readers
will be on allowing all of us to test either option _in Debian_. How can
we do that?

Given the far reaching nature of init systems in all system services,
the proper way to do that in the long run has been mentioned repeatedly
by Russ in the threads: *support in Policy* for *optional* upstart jobs
in packages. On that front, there seems to be quite some work done
already, at least for upstart (see #591791). People interested in these
discussion should really consider helping out policy finalization. It
will be way more productive than trying to win an argument before the
debian-devel audience --- which have close to no impact on the final
choice we will make. It is not clear to me the status of similar policy
work for systemd, although I see that systemd maintainers are
participating in #591791. Again, if you're interested in Debian switch
to systemd, please contribute to that work rather than arguing on

But given that no one is seriously thinking of making the change for
Wheezy, an important question is: how do we encourage more testing of
either options in time for Wheezy+1? I think it'd be great to have well
written guides that will allow Wheezy users to *experiment* with either
upstart and systemd. Similar documents exist in the respective packages
and have also been posted in form of blog posts on Planet Debian. People
interested in pushing for one of the two options, should consider
helping out with these documents. If they reach a good status, they can
also be proposed for inclusion in official Wheezy documentation. Nothing
like real feedback from our users will advance the cause of either
upstart or systemd in Debian.

Regarding porting, I recommend against using the argument "we should not
switch to something not supported by the $non_linux_port we released in
the past as technical preview". For one thing, the observation by
Christoph is very compelling (i.e. "first we choose, than we port ---
don't ask us to port before the choice"). For another, accepting that
argument will make us *more conservative* in the future about accepting
new ports. We will probably worry more and more about the impact on
(currently) popular ports, of accepting new ports as supported. In the
long run, we will probably diminish our willingness to accept new port
as supported. That would be a shame and also a strategic error.

Stefano Zacchiroli     zack@{upsilon.cc,pps.jussieu.fr,debian.org} . o .
Maître de conférences   ......   http://upsilon.cc/zack   ......   . . o
Debian Project Leader    .......   @zack on identi.ca   .......    o o o
« the first rule of tautology club is the first rule of tautology club »

Attachment: signature.asc
Description: Digital signature

Reply to: