Summary

  • Katsuyasu Toyama asked APNIC in August 2004 to establish an IPv6 Internet Routing Registry before wider deployment made routing records harder to manage. His draft left administration, architecture, maintenance and coordination open.
  • APNIC 18 supported the concept, then moved the details to a quiet mailing list. A planned April 2005 launch slipped during software customisation; APNIC records implementation on 11 May 2006.

The applause was the easy part. At APNIC 18 on 2 September 2004, a show of hands favoured Katsuyasu Toyama’s IPv6 Internet Routing Registry. Nobody objected. The chair immediately narrowed what the room had agreed to: the concept was accepted, while the design still belonged on the mailing list. He also remarked that the list was very quiet. A room had voted for a future service; the instructions for building it were still out in the hall.

Toyama's original proposal was unusually candid about the missing pieces. Who would administer the registry: a regional, national or local Internet registry? Should its database resemble the existing IPv4 IRR or another architecture? Who would keep records accurate? How would registries coordinate? He wanted APNIC to begin before IPv6 networks became widespread, because a correct routing registry could help operators check what neighbours announced. The conditional word mattered. A registry full of stale route objects would give IPv6 another address book to mistrust. He also named the bill: systems and operations cost money, and customers might pay more.

The meeting's answer was practical but incomplete. The service could be an extension of APNIC's IPv4 IRR if trust mechanisms proved sufficient. The chair sent framework and implementation detail to the list as action db-18-001. At APNIC 19 the following February, the Secretariat reported that it was extending the existing service, using RPSLng and changing attributes for IPv6. It expected the service by the end of April 2005. Toyama thanked APNIC for the progress. The action to discuss details on the list was still open.

April passed without the service. At APNIC 20 in September 2005, the Secretariat said customising the software had caused technical delays. Testing was under way. IPv6 operators were invited to register route objects on a test service before a production move. The same minutes mark db-18-001 done; that status closed the referral for discussion, not the production launch. A checked box on an action list was not yet a router operator's usable record.

APNIC's prop-025 history dates implementation to 11 May 2006, about thirteen months after the April target. It records a working capability, not proof that all RIRs adopted Toyama's hoped-for framework or that the new records were complete. The longer road also keeps the vote in proportion. Toyama asked for a registry while IPv6 still offered a chance to build record keeping early. The community agreed on the need. APNIC had to turn that agreement into software, testing and a service someone could actually update. The quiet mailing list was only one part of that labour; the official minutes name software customisation as the cause of the missed April target.