Chris, On Tuesday 25 January 2005 10:40, Chris H wrote:
Dylan wrote:
Since the mount point persists after the device/filesystem is unmounted, could a hardlink not be used?
No. Not for NFS as it will follow the link and read the original assigned device name.
There is no such thing as "following" a hard link. Symbolic links have direction to them. There is the symbolic link and the thing to which it refers. All hard links are co-equal names for an underlying file system entity and apart from the name itself, nothing distinguishes different hard links to a given entity. That said, other properties of hard links are relevant here. One is that hard links may not refer to directories (except for the one link that give the directory its name within its parent directory and the "." link that refers to itself). Second, hard links may not cross file system boundaries. Third, the mount point directory becomes inaccessible while a device is mounted on it. In effect, the mount point name changes the inode and resident device to which it refers when the mount operation takes place. Furthermore, file systems are devices mounted on directories. Only during the mount operation itself is the device name really relevant. Any use of a mounted file system (including sharing it via NFS) has nothing to do with the mounted device as such. That's the beauty of the Unix file system concept--it's a unified name space that can be dynamically extended by using the mount operation.
/ch
Randall Schulz