Re: [Ldap-dir] Concerns about draft-ietf-nfsv4-federated-fs-protocol

[Nicolas Williams <[e-mail]>]

On Fri, Sep 11, 2009 at 10:01:45AM +0200, Leif Johansson wrote:
> 
> > LDAP is supposed to ensure DN uniqueness atomically, is it not?
> >
> 
> Yes but only in a single operation. I suspect (but I may be wrong) that 
> several of the logical operations they need to perform would need to
> span multiple LDAP operations creating race conditions on the name.

I think they do not.

> > You'd create the FSLs first, then the creat/update FSNs to refer to the
> > FSLs.
> 
> That might work unless a semi-populated FSL entry is dangerous. 

An FSL alone is utterly useless and harmless.  An FSN alone too is
utterly useless.  It's only when junctions (which are local to servers)
refer to FSNs, which in turn refer to FSLs, that FSNs and FSLs matter.

Incidentaly the elephant in the I-D that wasn't specifically called out,
but which should have been, is that references to FSNs must be small and
fixed sized.  By using {UUID, NSDB hostname} they were able to keep
references very small: 16 + 256 bytes, or if you intern the NSDB
hostnames, possibly as small as 20 bytes.  This was crucial to my
understanding of the design :(