Re: [FedFS] meeting agenda (9/3)

On Thu, Sep 03, 2009 at 09:53:51AM -0400, James Lentini wrote:
> On Wed, 2 Sep 2009, Nicolas Williams wrote:
> > This being LDAP, I think a string with '/' will be much simpler.
> 
> Correct, but a string with '/' as the path delimiter is more 
> restrictive that what NFS allows for path components.

Understood.  You could also use the binary option and use NULs as
separators.  That'd be better, though it would make it harder to use
LDAP command-line tools.  Unfortunately you could not use multi-valued
attributes, as LDAP quite explicitly treats them as unordered sets
(RFC4511, section 4.1.7).

You could also have an object for every path name component, then chain
them together into paths.  That'd be overkill, and very hard to use.

> > I just checked with a local LDAP expert and he confirms my suspicion
> > that it's not correct to specify that o=fedfs (or whatever) be part of
> > the DN of every FedFS object.  The lack of search examples is definitely
> > a problem for me.
> 
> I would agree except for the fact that the NSDB is not a general 
> purpose LDAP database. The NSDB is an application specific directory 
> that stores only FedFS records (see discussion at the end of section 
> 4.1). For this reason, it is appropriate to use a well known root DN 
> to simplify the operation of the NSDB, fileservers, and administrative 
> tools.

It is one thing if a directory operator decides to segreate all NSDB
objects into o=fedfs.  It is another thing altogether for the WG to
decide it.  IMO it's inappropriate for the WG to make this decision.

Nico
--

-- 
_______________________________________________
nfsv4 mailing list
nfsv4 <at> ietf.org
https://www.ietf.org/mailman/listinfo/nfsv4