Re: Sense of the WG on DNS discovery

[Ralph Droms <rdroms@cisco.com>]

Eric,

As Rob pointed out, stateless DHCPv6 is described in
draft-ietf-dhc-dhcpv6-stateless-01.txt, which is in IETF last call.  There
has been one comment on the draft during the IETF last call,
http://www1.ietf.org/mail-archive/working-groups/dhcwg/current/msg02630.html.

The DNS configuration options are described in
draft-ietf-dhc-dhcpv6-opt-dnsconfig-04.txt, which has been approved as PS.

Several implementations of RFC3315, draft-ietf-dhc-dhcpv6-stateless-01.txt
and draft-ietf-dhc-dhcpv6-opt-dnsconfig-04.txt exist and have been tested
for interoperability earlier this year at TAHI, Connectathon and a DHCPv6
interoperability event hosted by NEC in Vienna.  Note that
draft-ietf-dhc-dhcpv6-stateless-01.txt describes a subset of RFC 3315, so a
server implementation of RFC 3315 would be interoperable with stateless
DHCPv6 client.  I contributed to the Cisco implementation of both a
stateless DHCPv6 client and server and use that experience as the basis for
my comments on the complexity of the implementation of stateless DHCPv6.

We discussed the issue of receiving multiple responses to the DNS
configuration information option in an ipv6 WG meeting.  Turns out, in
theory, the order of appearance of DNS recursive name servers in the list
doesn't matter because every server should return the same response to a
given query.  More generally, receiving multiple different responses from
DHCP servers (either IPv4 or IPv6) represents a misconfiguration and is an
operational problem - which could occur with other DNS configuration
mechanisms as well.

- Ralph

At 11:16 PM 11/6/2003 -0600, Eric A. Hall wrote:

>Rob Austein wrote:
>
> > - DHCPv6-Lite appears to have more support than the other alternatives.
> >
> > - DHCPv6-Lite does not appear to have any show-stopper problems.
>
>Is there an I-D describing DNS discovery over DHCPv6-lite yet?
>
>The last major discussion on this subject was in July. There was a lot of
>hand-waving about how easy it would be to use this technology for this
>service ("just do x, and y, and zz,"), but no effort has yet been made to
>my knowledge at actually deonstrating this simplicity with an I-D that
>actually describes the basic mechanics of the approach. I mean, there at
>least needs to be a discussion on client-side sorting and/or weighting
>algorithms for those cases when multiple answers are received, just so
>folks can see that this approach might actually work, or to get an idea at
>how difficult it will be. How can there be consensus on a proposal when
>said proposal does not exist?
>
> > The question to the WG here is not "Do you agree that DHCPv6-Lite is
> > the best choice?" (we know that some would disagree with that), but
> > rather "Do you agree that this is the sense of the WG?"
>
>There does not appear to be any such consensus.
>
>--
>Eric A. Hall                                        http://www.ehsco.com/
>Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/
>
>#----------------------------------------------------------------------
># To unsubscribe, send a message to <dnsop-request@cafax.se>.

#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.