To:
"Masataka Ohta" <mohta@necom830.hpcl.titech.ac.jp>, "Michael Richardson" <mcr@sandelman.ottawa.on.ca>
Cc:
<dnsop@cafax.se>
From:
"BELOEIL Luc FTRD/DMI/CAE" <luc.beloeil@francetelecom.com>
Date:
Mon, 28 Jul 2003 17:03:33 +0200
content-class:
urn:content-classes:message
Sender:
owner-dnsop@cafax.se
Thread-Index:
AcNUM/ArtkAq1MBpQMSdIT8I5CaEGwA3y6zg
Thread-Topic:
proposal for a compromise on DNS discovery
Subject:
RE: proposal for a compromise on DNS discovery
> > Daniel> Reality check: home networks are, if connected to the Internet, > > Daniel> attached via a gatewaying device. The de vice of choice for a > > Daniel> large number of networks is a $60 router from on of any number of > > Daniel> vendors. These boxes provide DHCP, DNS forwarding and > > Daniel> NAT. Perhaps some other environment would provide a better > > Daniel> strawman? DHCP is pretty prevalent in home networks. > > > > Right, exactly. > > > > I simply do not believe that there are environments where there will be > > a local (on wire) DNS server, but not a local DHCP server. The home > > environment is entirely the opposite - local DHCP server in SOHO router > > that gets the DNS info from the ISP via DHCP and/or PPP and passes it on via > > DHCP. > > Right. > > And, having such DHCP servers, there is no role of RA for stateless > (nor stateful) autoconfiguration, which is why IPv6 with ND is broken. > > The remaining problem is configuration on how to let the DHCP > server in SOHO router get the info and the solution should be > related to anycast. > > DNS server with anycast addresses does not neet intermediate DHCP. > > Masataka Ohta > Ok DHCP servers are embedded in current IPv4 SOHO router. This is cheap and efficient. But this is not a valid argument to to say that DHCP is THE only valid solution to announce DNS resolver server IPv6 address. SOHO routers, that are currently on market, are using a DHCPv6 client to be configured by the ISP but are using ND-based stateless address autoconfiguration to help local hosts to configure their IPv6 addresses ! Then what is(are) the solution(s): - RA-based - DHCPv6-lite - Anycast Addresses for resolver DNS server - SLP (I also like SLP, which already a RFC - Standards Track - !!! Does anybody have a good point against SLP ?) - well-known link-local addresses (+ DNS proxies...)?... There was a analysis made by the DNS Discovery Design Team in march 2001 (draft-ietf-ipnwg-dns-discovery-01.txt). Wouldn't it be valuable to go on or to restart such a work ? (instead of this no-end battle?) I must have missed something concerning this old work. Could anyone give me some pointers or summary? Luc #---------------------------------------------------------------------- # To unsubscribe, send a message to <dnsop-request@cafax.se>.