[Grml] Re: RFC: handling of external usb devices

[Mark]

> It fails as soon as multiple devices have the same fs-label.

Not as bad as /dev/XYZ breakage.  One can always construct failure
scenarios.  In all my years I have not once encountered a duplicate fs
label on any OS.  It doesn't happen.  Duplicate files happen all the
time, not duplicate volume names.

The problem is not bad anyway.  Designing a graceful error handler is
easy.  Pick one duplicate and leave the other unmounted.  Or do what
Linux does for files, append ".N" to the mount point name, where N is an
incrementing number.

> People don't read the docs (ask my mailbox). I want to avoid the use
> of bootoptions as far as possible.

I have never seen this thought expressed by grml anywhere.  If you have
a link or grml-tip or other command, I will read the docs.  The
important thing is not the handling mechanism, but this main point: OS
brittleness "sucks" a lot more than syntax.

> Yes, but UUIDs are long and IMO they suck a little bit on
> non-server-systems.

"They suck" is not a technical argument.  Mules and camels are uglier
than horses.  They also do far more work much more reliably.  UUIDs are
reliable.

> All symlinks will be deleted automatically as soon as the
> devices aren't present anymore. Nothing to care about.

OK; just think it through carefully with a USB-boot device that moves
from PC to PC to PC.  The BIOS can change hard drive ordering, etc.  At
next boot, all /dev entries will be wrong, so grml has to reconfigure
them gracefully.

> Usually you set up a swap partition because you want to *use* it. :)
> If you set up several swap partitions and want to use only a
> specific one I think you have to handle this in your own initscript.

Fair point.  The reason is tuning.  Tuning is important for large
programs with zillions of data elements.

It helps to assign swap to other drives and keep it off your OS disk. 
But you still want a swap on the OS in case there isn't any other swap
on the PC you're using.  Let alone cipher-swap issues.

           M