On 03/14/2013 11:31 AM, Jiri B wrote:
On Thu, Mar 14, 2013 at 12:17:50PM +0100, Peter N. M. Hansteen wrote:
On Thu, Mar 14, 2013 at 07:12:08AM -0400, Jiri B wrote:
just for curiosity, is it planned for future?
I can't just now think about real usability...
Me neither. For most use cases I can think of, interface groups (a feature we
do have, see ifconfig(8) and possibly other references elsewhere) will give
you what others have implemented interface renaming for.
Situation: onboard network card is broken and was used in OS.
You just plug additional network card, and disable the old
one via `config' (is this right?). The "policy" in your
setup is the order of network cards make some logic:
* 1st backup/installation
* 2nd service
* 3rd admin access
Now you don't use old broken card but you can't make new
one being first for example. I don't say this is good
design but I saw it used a lot in my previous job.
Renaming new card to old one is impossible.
disable the on-board card in BIOS, or since you obviously aren't
repairing the board, pry the chip off the mobo (yes, I've done
this...friend of mine gave me some re-badged Sokris 4501 machines with
bad NICs -- I popped off dead chip (it was the one getting too hot), and
suddenly my remaining ones became sis0 and sis1 (and the heat generation
dropped a lot). A little hot glue in the deactivated port, and I now
have a perfectly good 2 port Soekris.
But really...if you are living with dead on-board hardware, you need to
have the ability to make exceptions to policies like that...and in all
cases, some kind of labeling should be done.
[elsewhere in thread]
> So what is this good for in other OS?
...
other OSs have really stupid naming conventions.
They make up for the problems with their naming conventions by adding
"features". Those features create new problems, which are solved by
adding other features. Those features create new problems, so that
creates opportunity to make MORE features.
And everyone knows, the more "features" you have, the better it is,
right? The OS with the most features wins!
OpenBSD is for losers who actually have to get work done, not just
fiddle with time-saving features all day.
(yes, the default naming convention of OpenBSD causes some problems, but
they are easy to understand and easy to deal with. Certainly easier
than the "fixes" that try to eliminate dealing with the simple problems
by creating massive problems)
Me? If I have two identical machines with RAID 1 disks, and I have one
configured Just Like I Want It, I think I should be able to pull one
drive from the configured machine, pop both drives out of the second
machine, stick the removed disk from the configured machine in the
secondary, change the IP address and machine name, maybe remove the host
SSH keys, and be up and running. I should then be able to insert the
two free drives into the open slots and have the mirrors rebuild.
OR, if a machine fails and I have an identical machine, I should be able
to remove the disks, put them in the spare machine, plug the wires in
the same place in the spare machine, power on and be back in operation
with ZERO reconfiguration. This is something I should be able to walk a
non-technical person through over the phone (i.e., secretary, janitor.
Not managers, I have given up walking them through things).
This Just Works on OpenBSD. It doesn't work easily in most other OSs.
Nick.
