> There was a mode added to the 486 that allows it to run with a reverse =
> byte order (same byte order as the MC680xx).  This allowed NeXT to =

You appear to be misinformed.  The 486 (and later) have an
*instruction* called BSWAP which is a faster way to do what would
otherwise take a couple of instructions.  The performance improvement
is not all that much -- and not in any particularly critical areas --
leading people to suspect at the time that it was only used because it
guaranteed that the code would not run on a 386, where it was slow
enough to be unusable.

Back when MIT used "RVD" (Remote Virtual Disk -- basically a networked
block device not unlike Sun's ND -- though less convenient, if that's
possible) we had the problem of having two architectures, Vax and IBM
RT, both running 4.3 BSD and thus having byteswapped FFS structures.
This meant that RVD's couldn't be shared among architectures.  An
ffs-layer fix (I vaguely recall that it detected one of the magic
numbers, and if it had to be byte swapped, it turned on byte swapping
for everything else) was implemented shortly before RVD's were dropped
from widespread use, and I don't think it ever got deployed. Still, it
wasn't all that complex; the filesystem code is, after all, quite
modular and localized...
			_Mark_ <eichin@kitten.gen.ma.us>
			The Herd of Kittens