Re: Protocol: referrals and other URIs

["Jim Sermersheim" <jimse@novell.com>]

A couple IETF 58 attendees are proposing the following wording to
replace the existing wording:

Other kinds of URIs may be returned. The syntax and symantics of such
URIs is left to future specifications. Clients ignore URIs that they
don't support.

Jim


>>> "Jong" <jongchoi@OpenLDAP.org> 11/13/03 9:38:52 AM >>>
> >The server's knowledge on the other opaque protocols is an
implementational one
> >and hence needs to be treated as such in the protocol
specification.
>
> Yes, and as such, no server protocol implementation can be expected
to
> understand the relationship between operations and non-LDAP
protocols.
> And since no provision is given to allow an administrator to specify
> such a relationship, the current language is impossible to adhere
to.
>
> If there were some "applicable operations" field in referral URIs,
then
> the current language would be possible to adhere to. But there is
not.

The configuration of referrals in the server is implementation
depedent.
Should such relationships bw. operations and non-LDAP protocols
are needed, it can be specified in an implementation dependent way
without touching referral URIs. I agree to the need for a wording
change,
in that the server may return other kinds of URIs subject to its
knowledge
of that protocol, and the client may follow it also subject to its
knowledge.

> >The client needs to be prepared for the server which has imprecise
information
> >on the other protocols, but IMHO, this seems to be outside of the
protocol
specification.
> >- Jong
>
> I agree. Currently, there is no wording that assures a client that
it
> will receive a URI which it will know how to follow. This is good.