In your letter dated Tue, 3 Nov 2015 14:45:12 +0100 you wrote:
I'm not sure I understand why you are calling v6-only network 'broken one'? I've been sitting in v6-only network at RIPE meetings since that SSID was introduced. I'm in Japan now and so far I did not have to connect to dual-stack SSID - v6-only works just fine. What exactly (besides a few applications) is broken here?
(Can't wait for a request for Atlas to support NAT64, that's going to be interesting)
RIPE Atlas probe is a host. It should not care much if it's traffic is going via NAT64 box or not.
Let me start with the Atlas probe. When you schedule an IPv4 measurement the probe tries to use RFC-791 to perform the measurement. In an IPv6-only network that obviously doesn't work. So from that point of view, an IPv6-only network provides broken IPv4 connectivity. Another problem is of course that if you schedule an IPv6 measurement and pass the probe an IPv4 literal, then that obviously doesn't work either. If I have a device that does local DNSSEC validation (for example using getdns) then I lose access to the IPv4 world because DNS64 and local DNSSEC validation are not compatible. One area where I want to take a closer look some time is where NAT64 translators generate ICMPv6 packets that violate the specs. The reason they violate the spec is that the packets are converted IPv4 ICMPs. Some of this can be fixed by running 464XLAT on all devices. I consider that 'broken IPv4'. (And before we know it, the next fashion will be MAP-E and then all devices with have to do MAP-E as well, and MAP-T problably)