> Server location via anycast means that one is using the > routing system > to keep track of the location of the DNS server. That is, the > location of the DNS server is (somehow) injected into the routing > system, and the routing system is now responsible for > handing out that > information to any client that wants to know. If the server goes > down, moves, whatever, the DNS server becomes unreachable until the > routing system converges on a new useful answer,
=> Correct. This is the case for all traffic going through a certain 'path' in a network. and how > exactly does > the routing system know that the DNS server is gone, > anyway? Does the > DNS server now have to speak OSPFv3, or is there some human > sitting at > a 24x7 desk somewhere waiting to push a button on failure, or what? => There were several mechanisms shown in the analysis draft to allow this to happen without a human monitoring the wire. 1. The DNS can inject a route. Do you see problems with this? 2. Using Neighbour discovery and periodic solicitations. Most people in the DT didn't like this option. 3. The cleanest way: extensions to MLD to allow a node to join an anycast group. Hesham -------------------------------------------------------------------- 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 [EMAIL PROTECTED] --------------------------------------------------------------------
