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

Re: Ifupdown dysfunctional, is a Provides: interface possible please?


Actually, I would love to cleanly modify ifupdown with a lot of the
experience I have gained from netscript, and working on actual routers.
Making it not so onerous would not be that much work, and it would have
far better behaviour once improvements are made.

I am putting myself forward for this as with the systemd update
netscript-2.4 needs to be 'bodged' in a bit to make it work.  I will do
this, but getting ifupdown with a better operational method and feature
set will make it far easier to use for the server case as well.


Matt Grant

On Fri, 2014-03-28 at 15:08 -0700, Russ Allbery wrote:
> Josselin Mouette <joss@debian.org> writes:
> > Le vendredi 28 mars 2014 à 11:55 -0700, Russ Allbery a écrit :
> >> I realize that doing that well is not horribly challenging, but that is
> >> the most common server use case (and even desktop), and ifupdown does
> >> it quite well.
> > Come on. We all use ifupdown on our servers just because it is the
> > default and works well enough. Bringing up a pair of static IPs and a
> > bonding link is not very challenging. But we shouldn’t judge a network
> > management tool based on how to achieve the simplest task: any tool will
> > do that. Even Red Hat’s /etc/sysconfig/network-scripts, despite its
> > horrible design and antique syntax, works well enough for most of its
> > users.
> > And in the desktop case, I disagree that ifupdown does the job. User
> > applications want to be notified of the network’s status, and it
> > requires more than what ifupdown can offer.
> That sounds like vehement agreement with what I just said.  :)
> Having used both, ifupdown, despite its problems, is certainly better than
> the Red Hat approach in /etc/sysconfig.
> >> I don't want to lose that, and I don't want to add a bunch of
> >> complexity in order to satisfy that case.  I think there will always be
> >> a place for a very *simple* system to handle that case with some pre
> >> and post hooks for things like iptables rule installation.
> > This is one of the possible scopes of systemd-networkd. But I think it
> > is being designed more for cases like the initramfs, where you cannot
> > have a full-blown networking management tool like NM.
> It's quite possible that systemd-networkd will eventually become a good
> tool to do this.  It just isn't now.  Now would be a wonderful time for
> people who care to get involved, if they feel like that would be a good
> long-term solution, since it would be much easier to design a conversion
> path from existing ifupdown systems at this early stage in the project.
> -- 
> Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>

Reply to: