The concept of separation between file (inode) and name (directory) is
not new or unique to UNIX. I remember that MVS has this too: you
create a dataset (file) on a specified disk, and so that you don't
have to remember which disk that was, enter it into a system-wide
catalog (which I _think_ is just a designated file somewhere).
Uncataloged files are allowed, but on the installation I knew they got
deleted daily (i.e. for temp files).
The file management under MVS sucks unbelievably, but that's another
issue. ;-)
It boils down to the question: what is the canonical thing to access a
file? For MSDOS, SMB or ISO9660 filesystems it's a name, for UNIX-type
FSs or NFS it's a cookie (inode number) to which a name is simply
mapped. I think the cookie thing is more generic, but I'm not sure
which form of shoehorning is worse: applying the inode principle to
SMB or applying the name principle to NFS. (And I think the problem is
there whether it's a network or local FS; the possible races etc.
arise from multi-task capability already.)
Olaf
-
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/