>>>>> "BL" == Brett Lymn <blymn@awadi.com.au> writes:

BL> According to Gordon W. Ross:
>> 
>> I might suggest instead that strange formats like that might be
>> supported by a different ioctl that makes no attempt to hide the
>> attributes of the actual hardware (i.e. NEC 765 or Intel 82072)
>> and gives you pretty much raw access to that hardware.
>> 

BL> In that case, why not have an IOCTL that does a write track and the
BL> user provides a pointer to the buffer that has the contents of the
BL> track to write.  The IOCTL would have the assume that the caller knew
BL> what they were doing and had correctly laid out the track in memory.

Such an ioctl() would not be MI (in either Gordon's or Brett's case).
The nec765 floppy controller hardware wouldn't be able to handle Brett's
ioctl(), AFAIK, since it doesn't have the flexibility to do that sort of
format.  The kernel's driver would have to parse the user's buffer and
reverse-engineer the formatting parameters, hoping they fit something
that the hardware can handle--yucko!

==John