> One possibility that I don't think anyone has mentioned is
> giving the user or programmer finer-grained control over the
> name space for the unioned FS's. In other words, there could
> easily be times when you would want to make just certain files
> in the underlying FS take precedence in the name space.
One thing that sort of disturbs me ist that you'd really have a whole
filesystem that you'd have to mount and umount every time. I don't know
about POSIX standards etc, but to the layman (me ;) perhaps a kind of
'extended symlink' would look easier (to use, not necessarily to
implement...)
Take two directories, say /dir1 and /dir2, create an empty 'union mount
point' (mkdir /union) and join them:
ln -U /dir1 /dir2 /union
or perhaps with a different command:
union /dir1 /dir2 /union
Then all acesses to /union would get redirected to /dir1, if nothing was
found there, /dir2 would be searched. Write access would go to /dir1 (or
perhaps to somewhere else, whatever). You can, for static content, do this
now with a symlinking script. But not for dynamic content.
You wouldn't have any mount/umount nightmares, wouldn't have to care about
on what physical partitions your data is to 'unimount' them, it would be a
user space issue (at least, it looks like it ;) and all users could use
this feature, even if they did not have the possibility to mount/umount
things on a machine. It would not be an admin issue.
To me, this sounds like a Good Thing. I'd like to hear what you think,
especially concerning implementation.
-- _ciao, Jens_______________________________ http://www.pinguin.conetix.de- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/