Subject: Re: Item of future thought: disklabels
To: Brett Lymn <email@example.com>
From: Jonathan Stone <firstname.lastname@example.org>
Date: 07/25/1996 22:08:10
Brett Lymn writes:
>Why not just noffs (network order ffs)? That way you would not need
>to worry about be/le, you just mount it...
Because it's a subset of what was proposed.
ffsnb == ffsbe, by definition, and so adding ffsbe to a big-endian
machine is a no-op. If you have a native byteorder filesystem
on a LE machine, and the LE machine loses its magic smoke and dies,
having an "ffsbe" on a handy big-endian doesn't help you at all.
You need an ffsle to read the little-endian FFS from the dead machine.
ObHistory: NeXTStep/486 ran with a big-endian FFS on little-endian
machines precisely to support physically moving filesystems between
x86 and m6800 machines. Inlining the byteswap instruction in the
filesystem code was _the_ reason that NextSTEP/486 didn't run on 386es.