I'm going to stop cross-posting this :)

On Sat, 2002-03-23 at 20:11, Matthew Saltzman wrote:
> 
> I thought it would solve the problem to include the HWADDR parameter in
> ifcfg-eth0 and ifcfg-eth1, but that's only a partial solution.

Well, I'll be dipped; that's the first reference to HWADDR I've seen.  I
wonder when that was put in.  All the same, it looks like it might come
too late to be useful.  I'd appreciate any corrections to the following:

First of all, HWADDR is checked after the part of 'ifup' that checks
DEVICETYPE, so it'll never be checked for a wireless card... 'ifup' will
exec ifup-wireless.

It's not clear to me that switching the config in that manner would work
anyway...  All it does it call 'ifup' with the correct config file. 
However, that instance of 'ifup' may be unable to relate the config to
the logical link name.

Unless I'm missing something (could be... anyone get this better than I
do?) I still think that configuring the logical link names is a bad idea
and should be discarded.

> It seems that the interface names ethx are overloaded.  It also seems that
> the mechanism that is supposed to map interface names to fixed hardware
> addresses is broken as a result.  I'm not sure if the problem is
> initscripts or something more fundamental, though.

The problem can only really be fixed in the initscripts.  This wouldn't
be a concern if logical device names were always the same, but that's
only true until it's not.

> > Red Hat's system is much easier for them to maintain correctly with GUI
> > setup tools.  Rewriting network.opts would suck.  I'm all in favor of
> > keeping all of the information for a network card in one file, as Red
> > Hat does, but attaching the config to the logical link name, which may
> > vary, seems problematic for laptops especially.
> 
> I actually like the idea of the Red Hat system, but it's not working yet.
> This is one problem.  Apparently, another is making sure that the correct
> parameters are used when the wireless link is brought up and down.

The parameters should be correctly applied if the logical device names
match.  Mine works, when it does.

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to