In message <199908142013.WAA10929@orade.oreilly.de>Christoph Badura writes
>is@jocelyn.rhein.de (Ignatios Souvatzis) writes:

>Actually they are not at the start of a cylinder group.  This was done
>so that you could get a super block backups after a head crash.
>>you can 
>>resurrect at least part of your data even if the original superblock and
>>all but one of the cylinder groups are damaged.
>
>I've found that to not work too well in practice.

Odd; it works fine for many people.  Of course, like FFS, it predates
ZBR; the backup-superblock placement algorithm assumes fixed geometry
and spirals into the pack.  Back hten you taped the console output
(hardcopy) of the newfs output to the disk pack, right beside the
manufactuer's bad-block list.

I've fond use for superblock backups beyond the first even on modern
SCSI disks -- usually to recover from ghastly bugs in the pre-sysinst
i386 fdisk/disklabel mess, but also after other lossage.