Print

[normanr]

What about servers that provide multiple IPv6 addresses.  Pretty hard to check, but I guess testing that a DNS name has multiple IPv6 addresses, and all of them are reachable (via ICMP, or HTTP, or SMTP).

Also you'd have to check that the prefix was distinct enough, can't have just two addresses in the same /64.  What about support for 6to4? (How do you check the packets are routed correctly - i.e. not triangularly).

Also what about (and how to check) for a teredo relay server close to the server?

[snarked]

What about servers that provide multiple IPv6 addresses.  Pretty hard to check, but I guess testing that a DNS name has multiple IPv6 addresses, and all of them are reachable (via ICMP, or HTTP, or SMTP)....

Non-conclusive.  I have a physical machine that I use for virtual hosting and it has multiple IPv6 addresses (about 20), but each address maps to a distinct hostname - so there's no way to tell (vs. a common firewall/router).  Before I had the tunnel here, I used 6to4, and for a while, I had both the tunnel and 6to4 active in DNS.  However, I'm not certain what such a test for multiple IPv6s to the same host would prove - over separate implementations of connectivity.

Now, testing for multiple methods (tunnel-6in4, 6to4, native, teredo, etc.) might have merit in the short-term.  However, in the long term, every connection should eventually become native IPv6, so is this really something that should be certifiable?  It's a nice thought.