Hello Jan!

My vague idea about magical path beginnings to access underlying
volumes was not really though out, just seemed nice while I wrote.

As the decision about how to treat a certain mount point has to be
done for the mountpoint, not for the whole path (that may contain multiple
mountpoints) we would be forced into inserting additional levels into the
"special" paths at each mountpoint.

/coda/.MAGIC/mnt1/REGULAR/mnt2/REGULAR/mnt3/REPLICA0/file

Possible and theoretically clean (?) but not really convenient.

There exists a "real life proven" approach: DFS implements a magic
name (.rw) at each mount point despite that it breaks pure POSIX
semantics.

It is the one way to solve the naming problem - let us implement magic
names at each mount point. Not that I think it is perfect, but it is
rather easy to understand/use and hopefully not a nightmare to implement.

The deviation from pure filesystem semantics is negligible for practical
purposes, while of course this solution is in the same class of
hacks/workarounds as the Unix tradition to use "dotfiles"...

I think it would be acceptable to have something like (uppercase just to
denote the magical strings)

/coda/....../mnt/dir/file
/coda/....../mnt/.CODA_INTERFACE/REPLICA0/dir/file
/coda/....../mnt/.CODA_INTERFACE/BACKUP/dir/file

well, provided that the magical name (.CODA_INTERFACE above) is invisible
for readdir()...

[have browsed reiserfs4 notes (http://www.namesys.com/v4/v4.html)
Some of the ideas seem interesting, and the intention to clean up the
interface to file data and metadata is appealing.
They are kind of advocating special magic names everywhere.]

Best regards,
--
Ivan