Ooops, it gets rid of randomness *provided the same hardware is used*
but it should work if one changes slots etc.

On 28/09/2015, Edward Bartolo <edbarx@???> wrote:
> If I were to write this implementation I would do it this way:
> i) do the above as described to obtain a list of devices
> ii) sort the list alphabetically by name
> iii) rename the list as usual.
>
> That would get rid of all randomness.
>
>
> Edward
>
> On 28/09/2015, Rainer Weikusat <rainerweikusat@???> wrote:
>> fsmithred <fsmithred@???> writes:
>>> @Edward and David:
>>>
>>> Don't know if this is helpful or if you've already seen it...
>>>
>>> Here's how the new network interface naming scheme works (or is supposed
>>> to work):
>>> http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/
>>
>> Something worthy of being remembered for this case: This problem is
>> (almost) exclusively caused by the way 'distribution kernel
>> organization' ("compile every available driver as module") and "udev
>> module loading" interact: The kernel will probe devices as they're
>> encountered on the various busses but since udev loads driver modules
>> concurrently, this may even cause re-arrangements of "fixed" hardware as
>> the first driver which registers an interface gets eth0 (and so on), ie,
>> it's not the kernel names which are "unpredictable" but the order of
>> device driver init routine calls provided drivers are loaded by udev.
>>
>> Judging from the documentation, "the new scheme" "improves" upon this by
>> ensuring that not even single interface computers get stable interface
>> names: Move the card to a different PCI slot or reconfigure your USB -
>> hey presto! - your network interface just got a new name.
>> ?
>> ... and who wouldn't want his network interface to be named
>> "enp0s29u1u2"? After all, anybody unterstands the meaning of eth0 ---
>> how terribly boring!
>> _______________________________________________
>> Dng mailing list
>> Dng@???
>> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
>>
>