Speaker Angus Lees
Time 2009-01-20 11:40
Conference LCA2009

Team of 4 people.

We didn’t need permission, just do the work. Avoid management, avoid answer question: Why? Talk to technical people directly.

Separate duplicate network, so no risk to IPv4 infrastructure.

In Google, user demand drives everything. Get users first, then ask for permission.

Nothing new, nothing secret. Gain more by making a lot of noise, rather then keep it secret.

IPv6 statistics.

IPv6 not currently routed as optimally, in many cases, as IPv4.

0.238% users support IPv6.

0.09% users have broken IPv6 connectivity. That is users that won’t be able to access website with AAAA records. 36% of IPv6 users.

All browsers have own resolver libraries, may not prefer IPv6 address, depends on configuration.

Better IPv6 connectivity on Saturday, and most on Sunday.

Russia has biggest IPv6 penetration. Not Japan. Japan comes bottom of the list. Compared with IPv4 userbase.

  • 6to4 67.9%
  • Native/Other 29.1%
  • ISATAP 1.6%
  • Teredo 1.4%

We care more about “how bad is the brokenness” as opposed to “how fast is usage growing”.

You can’t ask user to choose IPv6. They shouldn’t have to choose, it should just work. If you have to ask, you have already lost.

Rather no IPv6 connectivity, rather then bad IPv6 connectivity.

Have had some success in forcing people to fix things up.


Test system only. Political statement. Not meant to be useful. Many people use it.

We hope to have an AAAA record in a year from now.

“IPv6 not inherently any less reliable than IPv6.”

Simple IPv6 traceroute can cross Atlantic several times. Issues slowly being reported and corrected as they are found.

Comparison of IPv6 with DTV. Not possible to turn off IPv4. May be possible to have mandatory IPv6 government requirements.