> No, they're both telling the truth, it's just that the truth that they're
> each telling isn't the same.
> 
> 96*1024*1024 = 100,663,296
> 
> A place I worked for a long time ago used 1024000 bytes as a disk "megabyte",
> chiefly because it made the examples in the manual for how to partition disks
> come out simple (2000 blocks was a megabyte, you see...).

One other thing to look out for is a situation like the following:

The boot messages for my machine include this
/netbsd: ncr0 targ 0 lun 0: <MICROP, 2210-09MQ1001901, HQ30> SCSI...
/netbsd: sd0 at scsibus0: 1008MB, 2372 cyl, 9 head, 96 sec, 512 bytes/sec

But my disk label includes this:
type: SCSI
disk: microp2210
bytes/sector: 512
sectors/track: 25
tracks/cylinder: 10
sectors/cylinder: 250
cylinders: 8261

A little work with bc says that:
2372*9*96  = 2049408
8261*25*10 = 2065250

So if I lay out my disk like it claims, I lose 8MB...  I'm actually
still losing a few sectors at the end of my disk, because the real
number of sectors is pretty close to a prime number.  (I don't
remember how many sectors the drive said it had, but it was not
divisible by 9!)  My guess is that the tracks have a variable number
of sectors >= 96, and they didn't want to lie about the number of
heads or cylinders, but I'm not a drive maker so I can't be sure...
Perhaps the IOmega people are doing this with the zip drive as
well.

Jon