>> But we can (and IMHO probably should) do better for either signle-disk
>> systems or for SCSI controllers, like Adaptecs, where two fields of
>> the C/H/S geometry is fixed for all disks on a given controller.

>I don't understand that last sentence...

[disklabels snipped]

By C/H/S I meant the BIOS geometry -- which AFAIK is fixed, at least
on old Adaptects like the 1542 or the Buslogic controllers we support.
(A fixed C/H/S geometry is not terribly meaningful on a modern SCSI
driver with ZBR.)

Just one more datapoint that confusion over what "Geometry"
means is rife :).