[RFC] plans for initscripts in F22

[Zbigniew Jędrzejewski-Szmek]

On Fri, Apr 25, 2014 at 01:30:00PM +0200, Lukáš Nykrýn wrote:
> Dne 25.4.2014 13:24, Reindl Harald napsal(a):
> >
> >
> >Am 25.04.2014 13:12, schrieb Lukáš Nykrýn:
> >>Dne 25.4.2014 12:50, Reindl Harald napsal(a):
> >>>Am 25.04.2014 12:40, schrieb Jóhann B. Guðmundsson:
> >>>>On 04/24/2014 04:30 PM, Miloslav Trmač wrote:
> >>>>>
> >>>>>Only those that are maintained directly inside Fedora.
> >>>>
> >>>>Which is what we care about we cannot hold back progress in the
> >>>>distribution based on someone, someplace, somewhere might be using
> >>>>legacy cruff.
> >>>
> >>>have you ever heard "if it ain't broken don't fix it"
> >>>network.service works fine until someone decides to break it intentionally
> >>>
> >>network initscript *is* broken
> >
> >no - such generalizations are always wrong
> >it does not fit for every setup and it don't pretend that
> >
> >proven by over 30 F19/F20 setups in a wide range from virtualized servers
> >with simple setups to physical hardware with multiple network cards, virtual
> >TAP devices acting as  routers, firewalls, WLAN accesspoints and VPN servers
> >with up to 5 decdicated openvpn-instances with their own keys, ports and
> >TAP devices it works for a lot of environments and they never will change
> >because that is why virtualization is used
> >
> >>During rhel7 beta I have discovered a lot of design flaws when people tried to use
> >>it on some advance hardware. Boot in fedora is now quite asynchronous and network
> >>is unable to cope with that. For example we have already removed the hotplug script.
> >
> >network.service is not for hotplug
> >it is for static configurations
> >
> >>And I really don't want to end with NM on laptops, network on simple servers
> >>and networkd elsewhere
> >
> >i really won't end with NM on simple virtual servers with one virtual NIC
> >so just don't break network.service intentionally because it does not fit
> >your usecases
> >
> >i don't demand you to you use network.service so don#t demand others
> >using NM and completly rebuild complex working setups - that's not
> >progress, that's just making development-noise to let people feel
> >there was done some work the hard way and they have to chew it
> >
> I agree. I also don't think that NM is the best solution for such
> use-cases. I believe that this is a place for networkd and I will
> not remove network initscript until networkd covers that.
Or even keep it around in its subpackage... maybe for a whole release
cycle. It doesn't really hold back other changes in any way.

Zbyszek