Re: alternate "dc" naming conventions

[Michael Helm <helm@fionn.es.net>]

"Kurt D. Zeilenga" writes:
> I thought your approach was, like the PKI suggestion, to use
> all the DC RDNs contained in the DN.

> >They're altogether in one sequence.  It's just the order
> >of rdn's is different than the locator draft's. 
> 
> So, your approach is to use the first (when looking right
> to left) adjacent set of RDN containing DCs.  This is interesting,

Yes

> but I'd still be concerned that DNs held in different servers
> would have to share the same SRV RRs.

Understood. 

Is there any reason to think this might not be a problem in
other infrastructures anyway (eg the Active Directory - based
W2K domain)?  Suppose a simpler example: one
directory that manages authentication, with basename ou=users,
dc=es,dc=net, & another that manages a phone book, basename
ou=staff,dc=es,dc=net; these two are only loosely coordinated
(or not).
Sure, the clients of these servers could get confused.
They would have to be smart enough to rotor thru the 
DNS SRV RR's they get back (unlikely).

> I believe co-existence requires the server which SRV points the
> client at to hold either the entry or knowledge information to
> where the entry is held.

Is it "requires"?  I'm hoping for that actually (this is my 
personal agenda, not nec. the grid community's), that the
grid directories would blend in to other available directories;
this is because nothing currently requires separating grid directories
from others, & separation (in the long run) places
an operational burden on the wrong people in my opinion.
But I think current trends suggest this "separate management" scenario
is what will happen as a side effect of a separate discovery
mechanism.

In the whitepages/auth dbms example above I would hope
that the two servers would give out knowledge references to
each other; I could use their DNS SRV entries and a little
indexing service to make sure that happened (convergence).
But would everyone want that?