Statistics abound on the deployment of IPv6 and where it might be in a couple years (hint: with a Logistic adoption curve, 50% of the world may be using IPv6 mid-2019).
So that’s the first half of the transition: dual stack. How are we doing on actually migrating to IPv6? What is the state of IPv6-only?
It’s not as crazy a question as it probably seems to those who haven’t started deploying IPv6. In the U.S., according to APNIC, some major networks are already 100% IPv6 enabled (or margin-of-error close).
We see here not just retail tech giants Facebook, Google, Microsoft, and Oracle, but universities like Virginia Tech and Rensselaer, cell giants Verizon, AT&T, and T-Mobile.
Independently, Apple’s Tommy Pauly provided this chart at a recent IETF:
Hurricane Electric keeps tabs on many things IPv6, including this table:
| IPv4 ASNs | 58054 | 99.1% |
| IPv6 ASNs | 13722 | 23.4% |
| IPv4-only ASNs | 44682 | 76.3% |
| IPv6-only ASNs | 350 | 0.60% |
| Dual-stack ASNs | 58605 | 22.8% |
A quick glance at many of those IPv6-only ASNs suggests that they have parallel IPv4-only ASNs. That further suggests that the networks are being built either separately, or modularly: the IPv6 network should not be dependent on the IPv4 network.
This is all consistent with initiatives at Microsoft, Cisco, Deutsche Telekom, T-Mobile, LinkedIn, and Facebook to make at least part of their networks IPv6-only. In most cases, there is translation for IPv4 at the edge, but the network is internally IPv6-only.
To the same end, I’ve begun working on Retevia, offering IPv4 translation as a service to IPv6 networks.
We no longer need to wonder whether IPv6-only networks will happen someday: they are already happening.