Re: init scripts don't fsck extra partitions

[Alfred M. Szmidt]

   - and not for 3: knowing what to mount where.

I think that 3 is actually useful, but for the human and not for the
command.  It is nice to have a list of what is being used where.

   I'd rather see /etc/fstab just got ridden of, and e2fsck run by
   /hurd/ext2fs itself upon startup (if necessary).

I don't like that.  For several reason, one is that you need to add
such special code into all translators, and don't have a generic way
of doing it for all file-systems.  Secondly, you'd need a way to
simply tell ext2fs (or any diskfs translator) to not run fsck.
Thirdly, e2fsck is specific to ext2, so you'd need to hard code which
translator to run for each specific file-system (you cannot trust the
name of the translator either).

I think that a far simpler, and more useful solution is to simply
generate fstab, and sync it to the disk when this translator exits.
It would simply check which partitions exist by probing once in a
while, maybe even once--say on install--since that info doesn't change
that often.  This could ofcourse be a translator. :-) Or simply a
script that does it, I think something like that exists for the
LiveCD's people have been making.

   Some advantage of this would be for instance to let the machine
   boot and administrator login while /home is still being checked
   [imagine more examples here]...  The fs could even be already
   mounted r/o while being checked (and ext2fs restarted if e2fsck
   made really big changes (nice equivalent to "please reboot" message
   on linux boxes)).

This has more to do with making our fsck smart enough to fork the real
file-system checker into the background for non-root-file-systems, and
then setting some flag that won't make the file-system that is being
checked mounted.