Re: dpkg semi-hijack - an announcement (also, triggers)
Petter Reinholdtsen wrote:
> I believe your understanding of Frans' position is fairly correct, but
> it differ from mine, and I am the one behind the dependency boot
> sequencing proposal. I am not convinced yet that triggers are needed,
> nor wanted, for dependency boot sequencing. This is partly because I
> believe the failure modes Frans mention in his post are unlikely to
> happen (or impossible, but it is harder to prove).
You are simplifying my reservations to a single point. My main point is that
insserv is just way too noisy in general, thereby making upgrades harder to
check by sysadmins and downright confusing to less experienced users.
insserv's failure modes are IMO extremely non-obvious.
> So those of you wanting to test dependency based boot sequencing do
> not have to wait for triggers for it to work properly. 80% of the
> packages got the dependency headers already, and those already using
> it report that it work quite well. :)
I do agree that more testing is needed and I'd love to see people basing
their opinions on their own experience instead of just trusting either
Petter or me. We are talking about a very fundamental function in any
installation and changing that in a fundamental way as insserv does deserves
a lot more care and review than it is currently getting.
Marc 'HE' Brockschmidt wrote:
> You seem to believe that Frans decides if and when we are going to
> switch to a dep-based init system.
That is a completely ridiculous simplification of Mike's mail and also does
not do justice to the concerns I raised. I have never said or even implied
that implementing insserv should be blocked just because of my concerns. I
have only pointed out my concerns to the RMs when I saw a request to elevate
something to a standard that IMHO is currently not up to Debian quality.
Have you actually tried insserv? Well, I _have_ tested it for about two
weeks on my laptop and basically had problems both with the switch and with
every upgrade involving init scripts after that.
That its maintainer requested elevation of insserv to standard just a few
weeks after he announces it ready for testing _and_ getting some non-trivial
negative feedback was rather an unpleasant surprise to me.
So, until you do try insserv yourself and can actually comment based on
experience, please don't dismiss concerns raised by others out of hand. Thanks.