Comment # 24 on bug 907457 from Jiri Srain

As discussed yesterday in Nuernberg: The situation when this situation happens
is rather rare (this is the only bug we are aware of). Since the importance of
the device map will decrease in the future (will be only written for legacy
booting), at the moment it is not worth any major changes. In particular:

- we will not introduce UUID for whole disks just for this purpose

- we could change the fallback for whole disks when UUID is not available. This
can differ based on the knowledge we have about failures of individual device
naming schemes, but the logic should be simple (e.g. use by-id unless running
on top of a specific hypervisor known to break by-id)


There is nothing we (YaST/installer) can do about making the naming schemes
stable under all circumstances (which would, admittedly, be the ideal
solution).