>Any reason why this is a bad idea? Any reason why sysinst couldn't just
>assume 1024/16/63 for the MBR on all machines with large disks (>503 MB)
>in case the user wants to use the entire disk for NetBSD? If he doesn't,
>sysinst could just grab the geometry from the existing MBR. In any case
>this stuff should be hidden from the user.

ummm because that geometry doesn't match that of the MBR ?

This matters when you use (p)fdisk to set up several partitions, one for
each operating-system. You want a translated geometry so OS partitions
after the physical 1024'th cylinder can be booted.

However one question I've had with translation:
  is there a standard way BIOS' do it?
  and if so, a standard that real-world BIOS do correctly ?

If not, what happens when I take a nicely setup disk and plug it into
another PC ? If the translation is done differently, the boot-loader simply
won't work.

And it's not like you can specify how the translation is done, at least in
the BIOS' I've seen. Sure you can enter the translated geometry sure, but
how do you tell the BIOS to divide the heads by say 5 and multiply
cylinders by corresponding amount ? I mean the BIOS could work it out by
dividing the BIOS heads by the physical heads and complain unless it was a
perfect divisor.

am I missing something ?

cheers,
Danny Thomas

sigh, why couldn't the BIOS API designer have used a full 16 bits for cylinder ?