> That's the problem: it depends on the processor's endianness and word size.
> Data is written on disk according to these processor characteristics...
Endian-ness, probably, but I don't think it depends on word size. Reiser
works on both i386 and alpha (from what I've heard), which would indicate
that it's not dependent on word size.
> I think so, but the ultimate goal would be to make it endianness and wordsize
> independent, like ext2. If we only fix reiserfs to work on sparc32, such an fs
> won't be readable on any other arch than sparc32. Not even sparc64.
A big-endian reiserfs should be readable on a sparc64; it should work just
as well as an i386-formatted FS on an alpha (though I don't have any actual
evidence showing that this works - anyone tried this?)
> Also consider the fact that we use a 32bit glibc on sparc64 because gcc
> miscompiles sparc64 code (apart from the kernel) for some obscure reasons. This
> should be fixed too...
Yes, but such fixes will be slow in coming, at least from me at any rate! I
can't afford any UltraSparcs. :-(
--
Mike "No .sig for you" Shuey
