> Note that such a program must be moderately clever in case a drive
> holding /usr moves from `box A' to `box B' where box B already has
> its own /usr.  (In the case of FFS, there are also endian-ness
> considerations, for drives moved from one machine to another.  The
> `last mounted on' field is a byte string and hence non-endian, so
> some other guard is required.)

I've been thinking about the endian issues, with the aim of someday
making it possable to swap drives from any one machine to another.
Admittedly there is a PhD and a few other projects in the to-do list
first.

My thought was to add two new filesystem specs: FS_BEBSDFFS and
FS_LEBSDFFS (or some such). Plus there'd be a specifically
big-endian and a specifically little-endian version of ffs. One
of the two would just map to ffs, possably maintianing a different
name just to reflect the specificity. The other version would
have byte-swapping macros where needed (All our ports are either
big endian or little endian, no?). It would mean a LOT of
gross swapping macros, but one source tree could probably handle
both versions. Then we could swap drives from any one machine to
any other. We could also build cross-endian miniroots. :-)

Note: I'm not proposing we do this too soon, but it's an idea.

Take care,

Bill