On Thu, 6 Mar 2003, Kurt Erik Lindqvist wrote:
I know that this has surfaced on the last RIPE meeting already, but I still think that a TLD name server is in no way special - it's IP address can be resolved by querying the root, so it doesn't need to have a fixed IP address "forever". If the address changes (which of course should not happen every couple of weeks) change the glue record in the root, and that's it.
Although I agree with you, reality is not always that simple. An advantage with locating a TLD DNS server at a IX point is that you have a easy way to get access to a large amount of carriers. We for example get transit from all members. If we today are to provide IPv6 connectivity, we are using the IPv6 space of our upstream provider. This forces us to rely on their routing policies, which might not be good from a "best for the Internet" view.
This said, the reason why we are having this debate at all is the fact that we don't have a PI mechanism, that said the reason we do not have a PI mechanism is the fact that IPv6 does not solve some of the largest problems on the Internet today. But that is a different discussion and I getting enough mail on other lists...:)
So - yes I agree with you, however this might lead to other unwanted problems.
This is why i said earlier i dont mind having a second "special case" on ccTLDs (being IXPs the first). Im seeing the "ok, then lets go to PI on v6 also" statement/resolution at the end of the tunnel -- and i dont really like it! :-) Making a "concession" on ccTLDs (giving them /48s) doesnt enable *any* company wanting PI to get such space. Because ccTLD management is well-known and it exists in a *limited* fashion. Regards, ./Carlos "Networking is fun!" -------------- [http://www.ip6.fccn.pt] http://www.fccn.pt <cfriacas@fccn.pt>, CMF8-RIPE, CF596-ARIN, Wide Area Network Workgroup F.C.C.N. - Fundacao para a Computacao Cientifica Nacional fax: +351 218472167