Re: /bin/sh (was Re: jessie release goals)
Le dimanche 12 mai 2013 à 19:40 +0200, Helmut Grohne a écrit :
> With all due respect, this might be utter bullshit, but is at least
> . I have yet to see a failing pid 1 (be that sysv,
> upstart or systemd). Acquiring data on failure modes of any of those
> systems appears like a difficult task and d-devel might not be the best
> place to discuss that.
Having a rock-stable PID 1 is nice and all, but it doesn’t help you if
something important crashes. On a production server, if apache crashes
and fails to reload properly because the scripts don’t get the ordering
right, it doesn’t help you that init is still running fine. It would
help you to have an init implementation that can detect which components
can be initialized and at what time.
We could buy a piece of the argument if systemd was actually prone to
crashing, but it is not. Most of the added features lie in other
binaries, not in PID 1 itself. Your system is much more likely to crash
because of a buggy driver in the kernel than because of the init system.
> The problem is not that people disagree on that a good init system is
> needed, but about what good comprises. Some people believe that a good
> init system should run on all supported architectures including
> kfreebsd-*. By this particular metric sysv init still outperforms
> systemd. In fact for every combination of init systems you will find a
> metric where one outperforms the other.
I was all for kfreebsd when it was proposed, but now that it exists and
nobody uses it, I am appalled at the idea of using it as an excuse to
stop making improvements to the linux ports. Should we stop any
migration to a decent networking system because BSD doesn’t support
netlink sockets, too?
.''`. Josselin Mouette
: :' :