To:
Alain Durand <Alain.Durand@Sun.COM>
Cc:
dnsop@cafax.se
From:
Jaehoon Jeong <jaehoon_paul@yahoo.com>
Date:
Wed, 16 Jul 2003 04:36:38 -0700 (PDT)
In-Reply-To:
<155F1C85-B75E-11D7-9A7A-00039358A080@sun.com>
Sender:
owner-dnsop@cafax.se
Subject:
Re: proposal for a compromise on DNS discovery
Thanks for your good proposal. As I said before, I believe, the RA-based and DHCP-based DNS discoveries are complementaty each other. One of them can not be the best. Just like IPv6 address autoconfiguration scenario, the RA-based and DHCP-based DNS discoveries can be used very well together. I think most of IPv6 hosts will have DHCPv6 client. Unless DNS information is piggybacked, IPv6 host just conforms to DHCPv6 procedure for DNS discovery. The approach of DHCP-lite announcing DNS information overrides IPv6 RA advertisement. IMHO, it is enough for one of RA and DHCP to announce IPv6 prefix and DNS information. In wireless networks, such as HMIPv6 and NEMO, MAP or MR announce IPv6 prefix information periodically. If piggybacking DNS information in RA, there is no need of additional delay for finding out DNS information. Your proposal complicates DHCP operation and usage, I think. DHCP is not almighty in all environments, and just one of many solutions. I am also one of non-english speakers. Thanks, Alain. /Jaehoon Paul --- Alain Durand <Alain.Durand@Sun.COM> wrote: > I have circulated this among several people > yesterday and it may be time > to bring this to the list. > > Our community is basically divided in two camps, > the "conservatives", coming from v4 where the > operational practice > is to use DHCP for DNS discovery and would like to > operate the same way > for v6, and the "liberals", coming from v6 where the > autoconfiguration > story is build on Neighbor Discovery and RA/RS, and > who would like to > expand on this. > > As I explain in an earlier mail, the differences in > the mode of > operation are: > - DHCPv6 can also pass the address of the NTP server > - RA/RS have the unsolicited multicast provision > that helps save > bandwidth > when a large number of nodes are autoconfiguring > at the same time. > > The compromise solution I'd like to bring forward is > to optionally > piggy back the > unsolicited multicast mode to DHCPv6-lite. > > More specifically, this would work the following > way: > > At boot time, the client wait a bit (random 0..MAX1 > seconds) and listen > to unsolicited DHCPv6 DNS recursive server > advertisements sent to the > link-local-all-node multicast address. > If noting comes, it send a DHCPv6 request. > The DHCPv6 server answer to the unicast address of > the requester > and, if configured to do so (this would be > optional), send the same > thing to the > link-local-all-node multicast address. > Those multicast responses will be rate limited by > the server, > that is, if a second request come within MAX2 > seconds, > only the unicast answer will be sent. > > I'm sure there would be many details to work out. > > - Alain. > > #---------------------------------------------------------------------- > # To unsubscribe, send a message to <dnsop-request@cafax.se>. __________________________________ Do you Yahoo!? SBC Yahoo! DSL - Now only $29.95 per month! http://sbc.yahoo.com #---------------------------------------------------------------------- # To unsubscribe, send a message to <dnsop-request@cafax.se>.