Summary

  • Randy Bush, Philip Smith and Jonny Martin authored APNIC's prop-062 in 2008. It reserved the registry's final /8 for one minimum-sized IPv4 allocation to each eligible new or existing LIR and held a /16 for unforeseen uses.
  • The proposal calculated 16,384 possible /22 allocations at the then-current minimum. Its authors also named the obvious loophole: a company could register several LIRs. At APNIC 31 in 2011, Bush defended the one-time rule and the option value of keeping addresses for networks yet to arrive.
  • APNIC began issuing from 103/8 in April 2011. The maximum later fell from /22 to /23. By October 2023 its staff counted just 121 /24s left in that particular block; recovered space formed another pool.

The latecomer's ticket

The consensus text of prop-062 reads like an attempt to save a seat at a packed table. APNIC was approaching the end of ordinary fresh IPv4 supply. Normal allocation rules could let a large, well-justified request swallow the remaining /8. Bush, Philip Smith and Jonny Martin proposed a different instruction for that block: divide it into small, once-only opportunities for both established and new local Internet registries. A network founded after exhaustion would still have a small public IPv4 foothold while building IPv6 service.

The three names matter. This was a joint proposal, and the archive records their collective design. It specified one allocation tied to APNIC's minimum size at the time of allocation, irrespective of the applicant's size or membership tier, after the ordinary eligibility criteria had been met. At the 2008 minimum of a /22, the authors calculated room for 16,384 such portions. They also put aside a /16 for unforeseen demands. In an address policy, that reserve is a little envelope marked “open when something surprising happens.”

The proposal supplied its own uncomfortable footnote. A determined organization might set up multiple LIR registrations and collect more than one portion. APNIC would need to watch for that, the authors wrote, while acknowledging the difficulty of perfect enforcement. A rule built around accounts can ration accounts cleanly; it needs a second kind of judgment to recognize common control behind several accounts. The loophole was visible at birth, right beside the arithmetic.

APNIC's proposal history shows the mechanism taking shape in public. Consensus came at APNIC 26 in August 2008. The allocation size was tied to the minimum in force, and the rule was detached from the fate of a separate global proposal. After final comment, the APNIC Executive Council endorsed it in November; implementation followed in February 2009. The policy existed before IANA handed APNIC the last block in 2011. Waiting until the shelf was bare would have turned design into a scramble.

Bush's ten-year problem

A policy meeting in Hong Kong in February 2011 made the personal reasoning clearer. The APNIC 31 transcript records Bush explaining the final-/8 design as one allocation per customer, once. He also described the operational bind. Small IPv4 pieces help future networks reach services that still speak IPv4, yet smaller pieces may swell the routing table. He speculated about a future of very small routed blocks, a prospect he said he disliked as someone whose work was to deliver packets. That was a forecast in a debate, not a measured routing outcome.

When another speaker questioned the value of sitting on IPv4 for years, Bush answered with a one-way-door argument. A later policy meeting could release a reserve once the need had passed. Addresses spent now could not be pulled back out of use when a new operator needed them. His comic alternative involved a donation to Greenpeace before proposing more IPv4 policy; the room laughed. The joke worked because the underlying allocation choice was painfully real. The meeting could revise text tomorrow. It could not replenish a block already consumed.

This was a distinct sort of scarcity management. The rule gave later entrants a chance to receive routable IPv4, which helped them meet an Internet still full of IPv4 peers. It also rationed expansion for an established operator with a demonstrated larger need. The applicants' identity and the registry's account boundaries became economically consequential. The same address could be a survival kit for a newcomer and an inadequate growth supply for an older network. The authors chose an entry path over a growth guarantee.

APNIC later reported that IANA delegated 103/8 on 3 February 2011 and that distribution from it began on 18 April. Each account could initially receive up to a /22. A 2019 policy change lowered the maximum to /23, stretching the pool. In October 2023, APNIC staff counted 121 /24s still available in 103/8 at the time of writing. They also described recycled IPv4 elsewhere in APNIC's holdings. The remaining count in one block was a snapshot; the policy's long life is the larger fact.

A small block with a large question attached

Bush's part of the story sits in two records: his name on the 2008 proposal and his voice in the 2011 argument over what to save. The documents do not turn him into the sole architect of APNIC policy. They show a coauthor of a precise rationing design who later defended its one-time limit and the practical value of an unspent reserve.

The design made a hard choice early. It protected the possibility that an organization founded years later could obtain a usable IPv4 entry point. It accepted account gaming as a policing problem and route fragmentation as an operating cost. Its success can be examined through later pool size, rule changes and who actually obtained allocations, rather than through the romance of the phrase “last /8.” The latecomer's ticket was never large. The decision to print any tickets at all shaped who could still join the table.

Source record