[gfs-wg] RNS critique

[Ted Anderson]

On 7/25/2005 21:00, Manuel Pereira wrote:
> ...
> As for the “alias”, this type of junction is really nothing more than a
> “referral junction”.  After discussing this with the GFS-WG, they feel
> strongly that “aliases” should be a basic feature of the service, however,
> agree that aliases (with hardlink “like” behavior) should be optional.

It remains to be determined exactly what hardlink-like behavior means.
There are two aspects of hardlinks that might be considered.  One is
that they are self-healing when namespace changes would break symbolic
links.  The other is that they are invisible, in the sense that any of
several names is an indistinguishable handle for the target object.

I have no objection to providing self-healing links within a repository
but I think the indistinguishable behavior would be a problem.  My
concern is that many of the objects linked to would be directories,
virtual directories in RNS or leaf referrals that would indicate
directories in another namespace.  Multiple indistinguishable links
means that directories would have multiple parents.  Thus paths for
objects would not be unique, so there would be no distinguished pathname
for objects in the namespace.  File systems have avoided this situation
for various reasons, allowing hardlinks only for non-directories.

Perhaps we should offer a junction with the self-healing property but
without making them indistinguishable from the primary parent-child
relationship.

Ted Anderson
IBM Almaden Research Center