> The first and most important rule of sharing a disk with MS-DOS and 
> NetBSD is this:  both OSs must agree on the geometry of the disk.  For 
> the sake of simplicity, I use the native geometry on the disk (which is 
> reported by NetBSD at boot).

And I, for similar reasons, use the fraudulent geometry enforced by the
Adaptec driver (32 sectors per track, 64 heads per track, for 1MB
cylinders), especially since the oh-so-clever BSD cylinder optimizations
were designed before drives started having three or four different
geometries in different areas of the drive.

Has anyone ever looked into having BSD (generically, not just NetBSD)
properly understand multiply-zoned disks?