Code review comment for lp:~raharper/curtin/trunk.net-manual

I'm not sure. My initial concern was if the file exists but was empty,
that systemd would not attempt to do persistent naming
as it expects that udev rules will handle it; then we don't do it, and the
system now has *no* persistent naming.

One could argue that's expected since user did not provide any matching
rules in the network.yaml.

If systemd applies persistent naming despite the presence of an empty udev
rules file, then that might make
the eni useless (iface eth0 does not exist).

However, if systemd doesn't do any naming because the file exists, then we
at least have a valid eni file despite
it not being persistent w.r.t which device shows up under the kname eth0.

I think I've convinced myself that we should at least ensure that systemd
doesn't do persistent naming. We could
emit a header in the file in case systemd checks for non-empty rules.

Shall I push an update or will you take in merge?

On Fri, Mar 11, 2016 at 1:17 PM, Scott Moser <email address hidden> wrote:

> It seems maybe arguable that you'd *want* to blank the udev rules if there
> were no interfaces to map.
> ie, if you're fully declaritive we own that file.
>
> thoughts?
> --
> https://code.launchpad.net/~raharper/curtin/trunk.net-manual/+merge/288799
> You are the owner of lp:~raharper/curtin/trunk.net-manual.
>