[ale] eth numbering change

[agcarver+ale at acarver.net (Alex Carver)]

On 2017-02-09 15:39, Chris Fowler wrote:
> 
> 
> --------------------------------------------------------------------------------
> 
>     *From: *"Alex Carver" <agcarver+ale at acarver.net>
>     *To: *ale at ale.org
>     *Sent: *Thursday, February 9, 2017 6:16:15 PM
>     *Subject: *Re: [ale] eth numbering change
> 
>     On 2017-02-09 14:33, DJ-Pfulio wrote:
>      >
> 
>      >
>      > On 02/09/2017 03:44 PM, Phil Turmel wrote:
>      >>>> Don't use ethX names with modern kernels.  Period.
> 
>     Ah, figures this was a systemd concept
> 
>     <ducks and runs>
> 
> 
> It does suck, but in this case I think they make a valid point.  I gripped about 
> this sometime back and then read up on it.
> 
> In my case I remove the config layer from the user where I install CentOS or 
> RHEL.  My config code is setup to look for eth0.   Those bits need to change to 
> those machines so that I know what LAN1 is.
> 
> USB is very unpredictable.  Many times I need to map a physical device with a 
> real config. Network interfaces allow me to use MAC.  Cheap USB->Serial cables 
> do not even have a serial # flashed so I have to resort to physical USB port to 
> tie a config to one.

I will grant that they have made an effort to also allow legacy behavior
(although they're also snarky with it so I don't mind being snarky
back).  This is a more friendly change than some of the other things
that have been done in the name of systemd.  However, with one exception
I can't recall really having an issue with network device placement.
They came up in bus order and had the same name unless I physically
reconfigured the hardware in which case I knew I was doing that and
simply made the alterations.

For USB I actually use just the bus chain as the identifier.  For things
that are left plugged in (most of my systems) it works pretty well.