I may have missed this in all the mail, but if you have sendfile() or
splice() and you have open-by-filehandle, then a user level NFS becomes
quite reasonable.
Is it open-by-inode that is the issue, or is it open-by-<anything which
is not a pathname> that is the issue?
Perhaps a better way to state the question is: is there any fixed size
handle that you would accept which would allow a user level NFS to
use the kernel's inode cache rather than reimplementing the cache in
user space?
For a lot of reasons, I like the user level NFS idea. It has problems
but so does the kernel level version. The user level version lets
people experiment with different file system ideas in user space and is
something I'd like to see supported. I want to use a modified version
of it to implement something very similar to clearcase views in bitkeeper.
-
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/