Summary

  • APNIC's 2004 customer-privacy policy removed default public access to member ISPs' customer assignment records. Toshiyuki Hosaka's January 2005 proposal sought published counts by country, address family and assignment size, with no customer identities.
  • The APNIC 19 Database SIG found consensus, the Executive Council endorsed it, and APNIC published its first dated assignment-statistics file on 30 August 2005. The meeting warned that APNIC's totals could omit assignments held by national Internet registries.

A privacy rule can close a window and still leave a question worth answering. Under APNIC's prop-007, member ISPs' customer assignments remained in the registry but ceased to be public by default in 2004. Researchers who had counted records through Whois lost that route. The change protected customer details; it also made the pace of address assignment harder to see from outside.

Toshiyuki Hosaka, then at JPNIC, did not ask to reopen the names. His prop-026 draft asked APNIC to count registered assignments and publish the totals by country and size. Allocation statistics showed which blocks had gone to resource holders. Hosaka wanted to know what those holders had assigned from the blocks. For an IPv6 policy debate, that distinction was the difference between space on a registry's books and recorded deployment beneath it.

The draft specified both assigned portable and assigned non-portable records, while excluding suballocations to avoid double counting. Its proposed line had a registry, country code, IPv4 or IPv6, prefix size and count. No customer name or contact was needed. He suggested at least monthly publication and left the final format to the Secretariat. This was a practical bargain: preserve the privacy boundary, publish the arithmetic.

At the APNIC 19 Database SIG on 24 February 2005, the Secretariat said it could implement the proposal. The chair found consensus after a show of hands. The minutes also identify the flaw that a tidy spreadsheet could conceal. National Internet registries held some assignment records, and their participation in equivalent reporting was their own choice. APNIC's published total could therefore be incomplete. Nobody in the room supplied a measured size for that missing part.

The Executive Council's 13 May minutes record no comments during the eight-week list period and endorse the consensus. APNIC's proposal history dates implementation to 30 August. Its archive for that year begins with assigned-apnic-20050830.gz. The original file contains 1,241 non-comment aggregate rows: 1,213 marked IPv4 and 28 marked IPv6. Those are rows in a report, not numbers of subscribers or assigned addresses. The deployed pipe-delimited layout also differs from Hosaka's five-field illustration. APNIC made the report real rather than treating the example as a frozen specification.

The first file answered one question and kept another open. APNIC could publish useful assignment counts without publishing customer identities. It could not, by counting its own records alone, guarantee a complete regional picture where national registries maintained other records. Hosaka recovered a public instrument for policy arguments. The instrument came with its coverage limit printed in the meeting minutes, even if a later table might tempt readers to forget it.