One more thought...

If you do break this hard dependency, I think that you will also
open up the possibility of creating the backupless file system that we 
have
suggested quite a while ago.

I'm not an expert on Coda, but I have worked quite a lot with AFS and 
DFS
and if Coda too has the possibility of creting multiple clones of a 
volume,
it would be possible devise a cloning scheme that would eliminate the
backup need created by the fact that users does the wrong thing, 
sometimes,
like rm * .o

One obstacle in creating such a sceme is the hard dependency that Ivan 
is talking
about. If this is removed, the whole management of backup volumes, 
including
their naming and mounting could be left to code outside of the file 
system, i.e. daemons,
cron jobs or whatever.

If the binding is preserved, the scheme would have to be implemented in 
the file system itself
probably with less flexibility, and with a larger burden of support for 
the coda group.

The other part, disaster backup could be managed through replication, 
now that disk
prices are approaching that of tape. This, I guess, is rather 
straightforward with Coda too.

Cheers.

Chalmers