On Thursday, April 4, 2002, at 07:45 AM, Ralph Droms wrote:
> I think the third scenario is likely to be problematic. > > My understanding is that your third scenario would > require that the customers' networks all be part of > the same site as the ISP network - at least as far in > as a DNS server somewhere. I imagine some ISPs > might want to draw a site boundary somewhere > between the ISP edge and the CPE; either in the > CPE or in the ISP edge device or by declaring the > CPE-PE link to be not in either site... This is the very reason why I suggested to use a well-known global Ipv6 address instead of a site local one. This address would have to be taken out of a reserved prefix that will not/should not be announced in BGP peerings. But in such a scenario, if prefix delegation from PE to CPE is done using DHCPv6, it would make sense to piggyback the address of the DNS server in this dialog and the CPE could act itself as a lightweight DHCPv6 server and pass the information to any internal node requesting the DNSserver IP address using DHCPv6... So, the question to ask is: In such a scenario, will it make sense to ask internal IPv6 nodes to implement a minimal DHCPv6 client or will they rather like to use a pre-configured well known address in their /etc/resolv.conf file? - Alain. -------------------------------------------------------------------- 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] --------------------------------------------------------------------