Summary

  • APNIC proposal prop-168, whose proposal text is signed by Christopher Hawker, converts a 3,013,120-address available pool into a 2035 exhaustion year; two links in that chain reproduce exactly from the proposal's own figures, and one does not.
  • The seven annual new-account-holder counts printed in the same paragraph average 770.7, not the 727 the argument uses, and the start date of the exhaustion clock is never printed — so the year, not the pool, is what a reader cannot check.

The sentence the argument rests on

On 29 July 2026, the proposal states, the APNIC available pool held 3,013,120 IPv4 addresses, or 11,770 /24s. That sentence is footnoted to APNIC's delegated statistics file for that date — a file the proposal cites and this reading did not open, so the pool figure is reported here as the proposer's citation rather than as an independent reading of the registry file (prop-168-v004).

Every quotation in this article is a provider-reported verbatim excerpt of an APNIC document, not a full-page rendering of it. That matters for the size of the claim: what follows checks the text that was printed, and nothing about page furniture, footnotes or later revisions that the excerpt did not carry.

From that pool figure the problem statement walks forward year by year: 549 new account holders in 2025, as of 03/12/2025; 667 in 2024; 904 in 2023; 824 in 2022; 783 in 2021; 827 in 2020; 841 in 2019. Then it states the conclusion plainly: "At the current average of 727 new accounts per year, and if each account holder was to apply for the maximum delegation of /23, the delegation rate would be approximately 1454 x /24 per year, meaning the pool will be exhausted in 2035."

Two links hold, exactly

The first is the pool conversion. 11,770 × 256 = 3,013,120. It is exact, and it costs a reader one multiplication. The arithmetic below is mine, labelled as mine, not the proposer's.

The second is the burn rate. 727 × 2 = 1,454. The doubling is the /23: two /24s per delegation at the current ceiling. It is exact, and it means the annual rate follows faithfully from the average the paper states.

That is the whole of the checkable machinery, and it is not trivial. Anyone can confirm that if 727 new holders each took a /23, APNIC would hand out about 1,454 /24s a year, and that a 11,770-block divided by that rate lasts rather less than a decade.

The third link does not arrive

The seven counts do not produce 727. They sum to 5,395, and 5,395 ÷ 7 = 770.7. The stated average sits 43.7 below the mean of the very numbers printed beside it.

The paper does not say what 727 is. A weighted average would do it; so would a different set of years; so would a correction the text does not show. Two features of the list make a plain mean of it awkward in any case. The 2019 count begins only with prop-127's implementation on 04 April 2019, so it covers part of a year. The 2025 count stops on 03/12/2025, so it also covers part of a year. Averaging seven years when two of them are stumps is a choice, and the choice is not described.

The proposal is candid about the second stump: it notes that full-year 2025 new-account-holder statistics were not published in APNIC's 2025 Annual Report. That candour is worth naming, because it is the reason a fuller recomputation was not available to the proposer either.

The year is underdetermined twice over

Take 1,454 /24s a year against 11,770 and the pool lasts 8.10 years. Counted from 29 July 2026, 8.10 years lands in 2034, not 2035. To reach 2035 on that rate the clock has to start at the beginning of 2027 — a date the proposal does not print.

Move the average and the year moves with it. At the printed-count mean of 770.7 the pool lasts 7.64 years, ending early in 2034. At the latest full-year count the paper prints, 667, it lasts 8.82 years, ending in the middle of 2035. The band between those two computations is more than a year wide, and both ends are built from figures in the same paragraph.

None of this makes the pool healthy. Every version of the computation puts exhaustion inside the coming decade. What it makes is the specific year — the one printed in the sentence that carries the argument for changing policy — a conditional output, sensitive to an unprinted weighting and an unprinted start.

The assumption is stated, which is a point in its favour

"If each account holder was to apply for the maximum delegation of /23" is written into the sentence. That is not evasion; it is a stated ceiling. It should be read as one. A rate that assumes every new holder takes the largest delegation available is an upper bound on delegation demand and therefore a lower bound on how long the pool can last. The proposal's own framing — resources sitting idle "up to 9 years" — follows from that ceiling, not from observed take-up.

Which raises the awkward part. The ceiling is the current /23 maximum. The change the proposal recommends — allowing a sub-set of account holders holding less than an aggregated /22 to receive an additional delegation up to a /22 (prop-168 page) — would raise that ceiling. So the exhaustion rate in the paper describes the world before the proposal, and no revised rate is printed for the world after it. The date is the last one computed on the old maximum.

Two surfaces, two statuses

The proposal page prints its current status as "For discussion at APNIC 62 OPM". The proposals index files prop-168 under "For Discussion on Policy SIG Mailing List" (proposals index). Both readings are of APNIC's own pages, on the same day, and they do not say the same thing. A stale page is the obvious explanation and cannot be confirmed from the excerpts; neither surface is treated here as the governing one.

What a reader can still check

Holding only APNIC's publications, a reader can confirm the pool conversion, confirm the doubling, and derive about eight years of burn at the stated rate. A reader cannot reproduce the average, cannot see how two partial years were handled, and cannot learn when the clock starts. The chain is exact at both ends and missing its middle.

Closing that gap is cheap: the formula behind 727, the treatment of 2019 and 2025, and a start date. Three lines of footnote would turn a persuasive number into a checkable one, and would let a voter separate the arithmetic from the assumption rather than having to accept both together.

A proposal that argues the remaining life of a shared pool owes the people voting on it exactly that: the pool, the denominator, the formula, the start. Prop-168 supplies the first two. The proposal may well be right that the pool empties within a decade. Its own figures say something narrower and more defensible — roughly eight years of burn at the current ceiling, conditional on maximum take-up, with one unprinted weighting moving the end date by more than a year either way. That is enough to justify a debate. It is not enough to date the end.