On Tue, Mar 15, 2005 at 07:00:36PM +0100, Guus Sliepen wrote:
> On Tue, Mar 15, 2005 at 05:41:11PM +0000, [EMAIL PROTECTED] wrote:
>
> > > However, I just thought
> > > that it might be even better to create an if-pre-up.d script that
> > > unconditionally runs ifrename for each ifup command.
> >
> > Are you suggesting that
> >
> > ifup foo
> >
> > do this?:
> >
> > ifrename -i foo # Renames foo to bar
> > ifup bar
>
> No, I'm suggesting that
>
> ifup foo
>
> does this:
>
> ifrename # Renames eth7 to foo
> ...do rest of ifup stuff...
>
> At boot time, the effect is the same as an S40ifrename script. However,
> this way it also works if you ifup at a later time and there are new
> network devices.
No. If you have 3 interfaces, you will call ifrename 3 times
instead of one, and each time ifrename will have to probe all 3
interfaces, so it's going to take more time. I personally do care
about boot time (which is why ifrename is designed the way it is and
why I'm going to do something about hotplug/net.agent).
> > I don't see the point in supporting hot plugging on systems that don't
> > have the hotplug package installed.
>
> Well, I used to do that a lot, mainly because all the hotplug solutions
> sucked.
Unfortunately ignoring it will not make it go away (like the
previous maintainer of ifupdown did believe). For USB and CardBus,
Hotplug is the *only* solution. And 16 bits Pcmcia will migrate to
Hotplug by the end of the year (i.e. cardmgr and its scripts are
dead).
I was in your case. I did my list of things that I though
sucked in hotplug, and I started fixing them. Things are not perfect,
but I'm sure Thomas will acknowledge that we made progress. It's just
a shame that all the work of Thomas is stuck in limbo because of the
freeze for Sarge.
Please make your own list about what's wrong in Hotplug (and
don't forget to read my document). Then, we can discuss the individual
points.
> > Roaming is another issue. When someone moves from one location to
> > another he can reconfigure an interface; however, he doesn't normally
> > _rename_ the interface at that time.
>
> True.
Maybe I should allow to ifrename based on IP address ;-)
Jean
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
