Adam Megacz wrote:
> Russ Allbery <[EMAIL PROTECTED]> writes:
>>>   2. In some alternate universe, would it have been possible to
>>>      implement mountpoints as symlinks to some volume-name space?  For
>>>      example, if the cache manager automatically mapped
>>>      /afs/cell/@vol/volname to the root directory of volume "volname"
>>>      in cell "cell", mountpoints could be implemented as symlinks.
> 
>> Mount points *are* symlinks.  So I'm not quite sure what advantage this
>> would have over the current state.
> 
> True, but the cache manager grants them special semantics above and
> beyond the normal semantics of symlinks.  I'm curious about what
> advantage that special treatment brings.
> 
> Is there any particular use-case that called for the
> "xyz/mountpoint/../ == xyz" property which normal symlinks do not
> have?  Are there any other properties mountpoints have which other
> symlinks do not?

A mount point is implemented in Posix as a symlink.  It is not a symlink.

A mount point points to a directory that does not have a parent.  That
is a very significant difference.  The behavior of the cache manager
takes that into account when handling mount points.

Jeffrey Altman

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to