Rémi Denis-Courmont (m/j/a) 10/6/08 5:55 PM:
Hello, Le lundi 6 octobre 2008 17:08:01 Rémi Després, vous avez écrit :On 3/10/2008, at 2:26 AM, Erik Kline wrote:Well it seems to be working well enough (or popular enough, anyway). What if I said that 69.8% of IPv6-enabled users who visit google.com have 6to4?As we discussed in Montreal, this proves that there is, for v6 packets that have the prefix of 6to4 (2002::/16), a path from ipv6.google.com servers to the IPv4 Internet. But it doesn't prove that clients of IPv6.Google, if they try to reach another IPv6 server, will always have a 6to4 return path for their packets. similarly, they may not be reachable, as servers, from any client having native IPv6 connectivity.Last time I checked, Google used HTTP, HTTP ran on TCP, and TCP checked return-routability in the three-way handshake.
Did I suggest the contrary in any way?
I don't disagree that the lack of built-in return-routability in 6to4 _is_ a problem in certain circumstances. Namely, when there is an IPv4 firewall on the proto-41 path. But lets not pretend it is worse than it really is.
1. In my understanding, the problem is not only with firewalls. If you have seen a proof that all ISPs that offer IPv6 have a guaranteed route for the 2002::/16 prefix toward the IPv4 Internet, with a quality they can control, that's good news :-), but it's not the state of the art I know. 2. I don't "pretend" anything. I just tell what I believe is true, when IMHO it helps IETF to make progress. Please remember that thanks to this kind of analysis, I now have native IPv6 at home. My ISP, Free, which didn't consider offering IPv6 across 6to4 relays, deployed IPv6 at no charge for customers after I had explained 6rd's potential. Regards, RD