Yep, names like p4p1 are soooo much better than eth0 :-(

[agraham]

The real problem here is the designers of the concept lost sight of the 
actual benefit to the user, the problem as I would state is:

"Provide a means that allows consistent naming of network devices".

That should have translated into eth0 is "ALWAYS" the first device, eth1 
is "ALWAYS" the second device etc.. the biosdevname should have then 
been used to create that relationship and _nothing else_.

If the device does not have the BIOS entries, simply fall back to the 
existing scheme.

The above would have provided the "consistency" and "compatibility" I 
think everyone was actually looking for and no one would even need to be 
aware of the feature.

IMO this feature needs to be re-worked to meet this simple design goal 
that will then benefit everyone, especially RH et al. because in the 
real world, in any enterprise environment, this feature will be turned 
off immediately due to the confusion it causes and no-one benefits.

Albert





On 10/13/2011 01:23 AM, Tom Horsley wrote:
> Especially when a new biosdevname package is installed and it
> decides the name is no longer p4p1, but is now p6p1.
>
> All the iptables rules that refer to the interface name
> are broken.
>
> The ifcfg-p4p1 file needs to be renamed and edited.
>
> Wasn't the theory propounded that these new names
> would cause less confusion?
>
> (And why was having some disks named hda0 versus sda0
> confusing, so we had to change all the names to sda0?
> Can these two theories be made compatible? :-).