On 7/29/2017 12:42 AM, Randy Bush wrote: > so folk have some data to consider, as of tuesday about 08:00 gmt we had > seen the following unique associations (ietf-hotel excluded, as is not a > measure of choice) > > 1380 ietf > 967 ietf-legacy99 > 611 eduroam > 226 ietf-2.4ONLY > 166 ietf-nat64 > 62 ietf-v6ONLY > > of course, a device could have been a serial associator, using more than > one network over the measured time period. I may be a serial associator, since I connected my laptop to ietf, ietf-nat64, and ietf-v6ONLY. The connection to "ietf" was automatic, something the laptop remembered from time past. I purposely tried ietf-v6only, because I like try stuff. Did not work well. One of the servers I rely on is IPv4 only, and that was just not accessible. I fell back to ietf-nat64, and it was OK. So if the purpose of "eating our own dogfood" is to find out bugs and breakage, here is a clear finding: IPv6 only networks do not provide access to IPv4 only services today. NAT64 networks seem to work, but we suspect that there are some drawbacks, such as messing up with DNSSEC, and maybe also interfering with some P2P services. So here I go with my $0.02 suggestion: moving the default to IPv6-only won't work, but moving to NAT64 would mostly work. It would also probably teach us what exactly is wrong with that technology. -- Christian Huitema