[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Proposed IPv6 DNS Discovery Requirements



    Date:        Mon, 13 May 2002 19:15:46 +0700
    From:        Robert Elz <kre@munnari.OZ.AU>
    Message-ID:  <4427.1021292146@munnari.OZ.AU>

  | How can this work if the server isn't in the same site as the client?

In reply to this query of mine, a private message (thanks for that)
reminded me of a message that Itojun sent to the list back in early April...

  | Message-id: <2012.1017973554@itojun.org>
  | To: Ralph Droms <rdroms@cisco.com>
  | Cc: ipng@sunroof.eng.sun.com
  | From: itojun@iijlab.net
  | Subject: Re: Stateless DNS discovery draft
  | Date: Fri, 05 Apr 2002 11:25:54 +0900

  | 	if CPE can become dual-sited (participate into ISP's site and
  | 	customer's site) it can relay DNS query requests/responses between
  | 	clients in customer's site to DNS server in the ISP.


which is true, messages could be relayed - however I believe that
breaks the "serverless" requirement that is being suggested.

That is, to relay a site local message effectively, the reply
also needs to be relayed (the message into the ISP needs to be
sent from the relay's address inside the ISP's address space, as
the customer's site local address is useless there).  So the
reply will return to the CPE, and then needs to be relayed back
to the end node (with the appropriate address substitutions made)

To accomplish that, either the relay needs to retain state
(it would be close enough to a specialised NAT server) or the
protocol needs to include enough information so the relay can
tell from the reply where the reply needs to be sent (which
makes the whole protocol close enough to isomorphic to DHCP,
and certainly could not just be DNS packets to a well known
address).

Either way, I don't think this meets the objective.

Aside from the "we aren't sure deployment will be good enough"
is there some reason why multicast isn't being used for this
search?   Or more bluntly perhaps, why svrloc isn't just being
used?

Multicast deployment will follow closely upon a requirement for
multicast deployment - as long as no-one wants to use it
because it isn't deployed, it never will be.   Require it
for a worthwhile application, and deployment will simply happen.

kre



--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to majordomo@sunroof.eng.sun.com
--------------------------------------------------------------------