According to Frank van der Linden:
> On Mon, Apr 20, 1998 at 07:58:42PM +0200, Lucio de Re wrote:
> > [ ... ]
> > 
> > 	<filename>: Stale NFS file handle
> > 
> > [ ... ]
> 
> In normal operation, this error can occur when a client uses a handle
> that is no longer valid, usually because the file it references
> has been removed on the server.
> 
I've done some digging, and at least ascertained that the NFS server in 
Plan 9 generates the error stale handle condition.  However, this is 
caused by the RPC credentials for the client having a zero length 
machine name, and I'm not sure how to address this aspect of the 
problem.  I am expecting some replies from the Plan 9 fans mailing 
list, but it seems that NetBSD opinion should be consulted as well.

The RPC rfc (1057 - June 1988 :-)  does not seem to prohibit this 
behaviour, so I'd like to hear from the implementers whether there is 
any mechanism to work around it.

As a second gripe, I note that mount_nfs in versions 1.2.1 and 1.3 
assumes (understandably) a version 3 NFS server.  The Plan 9 server is 
version 2 and this caused the mount never to succeed.  Something seems 
ripe for correction in this: either the portmapper on the other side 
somehow responds with a demoted version (I think the RPC rfc forbids 
this) or the mount client takes it upon itself to demote its request.  
The -2 and -3 options can then be treated as obligatory.  This was my 
assumption of mount_nfs's behaviour, but it seems my assumption was 
mitstaken.
-- 
Lucio de Re (lucio@proxima.alt.za)
Disclaimer: I'm working at getting my opinions to agree with me.