Our IPv6 implementation plan was mentioned in Circle ID ( http://www.circleid.com/posts/20090309_slow_mainstreaming_of_ipv6/) yesterday. We’re proud to report that there is some progress on rolling out IPv6 but it’s just that so few people care.
We’ve been working on IPv6 for at least a few years. Last year was the first year that we turned up any IPv6 to public facing services. While you can use our Custom DNS service with IPv6, I cannot say that we have had overwhelming demand for it. “Connect my IPv6 to the Internet” is what we typically hear. The joke shows that users just exposed to IPv6 think of IPv6 as a more difficult Internet to manage. Have you seen reverse DNS entries in IPv6?
Until the choice of IPv6 versus IPv4 is made on behalf of users, the choice will always be IPv4. It’s more familiar, easier, and it just plain works. Getting someone to choose IPv6 over IPv4 is fighting the success and network effect of IPv4.
There are a number of challenges we have come up against. One is on routing. You can also see our IPv6 announcements in the CIDR Report for 33517. This little victory underscores the difficulty RIRs have with the concept of anycast. Specifically that routing policy and allocation policy are evaluated in two black boxes. (And no, we will not suggest a change to the mailing list.)
Another is how recursive DNS resolvers respond to queries when there is IPv4 or IPv6 transit and whether a client queries with AAAA or A. This has caused a load of AAAA queries that often outnumbers the number of A requests. How this behavior will shake out as more user IPv6 transit becomes available will be interesting to watch.
Another is how IPv6 nameservers have to be added to EVERY single registry. I understand that it’s good practice to only allow valid nameservers but it’s still a pain when we’re turning up domains in every ccTLD (just did .fi or .de. recently). It reminds us of the pre-activation woos we had to go through. According to the RFCs, you are not to answer authoritatively for a zone unless it was delegated to you. Well, some registries would not create a lame delegation by adding the NS entries before a nameserver was authoritative – creating a chicken/egg problem. In the grand scheme of things, lame delegations are probably worse but it was still another process to investigate with various registries.
There is still a long way to go but it’s getting there.