Summary

  • Prop-168-v003 would let eligible account holders below an aggregate /22 apply for more IPv4 space up to that ceiling, but needs assessment and finite supply mean the ceiling is not an allocation guarantee.
  • The 18 August impact assessment counted 11,729 /24 units on 13 August. Reserving one /12, or 4,096 /24 units, would leave 7,633 for ordinary distribution—insufficient if every potentially eligible APNIC and NIR member sought the maximum.
  • The proposal also creates a likely application queue outside the standard delegation SLA and a five-year transfer clock whose scope across older direct allocations and M&A needs clarification before APNIC 62.

The offer contains a subtraction

Prop-168 is usually introduced by comparing two prefix lengths. Current policy limits a direct delegation to /23; version 3 would allow an eligible account holder with less than an aggregate /22 to apply for enough additional space to reach /22. That sounds like a doubling from 512 to 1,024 addresses.

The operational story begins elsewhere. A /22 is four /24 units. A /12 is 4,096 of them. The Secretariat's 18 August assessment said APNIC held 11,729 /24 units as of 13 August. Creating the proposed transition reserve therefore leaves 7,633 /24 units in the ordinary pool.

That remainder matters more than the headline ceiling. The assessment says it would not cover the maximum requests of every potentially eligible direct APNIC and National Internet Registry member. Applicants would still need to pass a needs assessment, and not everyone will apply. The document does not predict actual demand. It does establish that the rule cannot promise every eligible account the advertised maximum.

The reservation also has to occur before ordinary exhaustion. The proposal describes a /12 for new account holders that need a /24 while deploying IPv6 transition mechanisms after the general pool has run out. But a registry cannot wait until there is nothing left and then set 4,096 units aside. If the proposal is adopted, the protected stock has to be marked at implementation.

This is not an argument against the reserve. Preserving small IPv4 blocks for later IPv6-transition use can protect entry after the ordinary pool closes. It is an argument for naming the subtraction at the same time as the benefit. The general pool and the transition pool serve different recipients at different moments; counting them together conceals that choice.

Eligibility becomes a place in line

The proposal does not distribute space merely because an account falls below /22. An applicant would submit a request, document need and pass assessment. Account holders that transferred IPv4 space out would be ineligible. If demand arrives in a burst, the order and speed of those decisions become part of the resource rule.

The impact assessment expects high application volume and significant processing delays. More unusually, it says delegations under the proposal would not be covered by APNIC's standard delegation service-level framework. Helpdesk service would retain its ordinary targets, but the scarce decision itself would not carry the same clock.

That distinction changes the meaning of access. A technically eligible operator may need addresses for a network launch, a customer migration or additional translation capacity by a particular date. Approval after that date is not equivalent to approval before it. The operator may still have to buy, lease, renumber, delay deployment or add carrier-grade NAT while its request waits.

The archived proposal text already anticipated a sudden rush and extended assessment times. Version 3 now sits beside a quantified remainder. What is still missing from the public decision is the ordering rule: timestamp, completed evidence, lottery, proportional ration, NIR channel, or another method. “A waiting list may return” is not yet a complete allocation algorithm.

The issue is especially important because the scarce unit is not just an address. BTW's history of the 103/8 ration shows that a quota attached to an account changes incentives around membership, corporate structure and transfers. A queue attached to accounts inherits the same problem: the public needs to know whether it is counting legal accounts, independent networks, controllers or operational projects.

One new block may reset an older clock

Prop-168 also asks an applicant to exchange mobility for direct supply. An account that has transferred any IPv4 space out would not qualify for more. After a new direct delegation, available-pool space could not be transferred for five years from the most recent delegation.

The difficult words are “most recent.” If an account received direct blocks at different times, the proposed text appears capable of restarting the clock across older direct delegations when a new one arrives. The assessment also says the wording appears to reach both ordinary market transfers and transfers associated with mergers and acquisitions. It records a conflict between the text and examples discussed by the author and asks for clarification.

That is not a drafting footnote. A small network might welcome several hundred lower-cost addresses while also expecting a financing, sale, merger or restructuring within five years. The value of the new block depends partly on whether it freezes only itself or alters the mobility of older holdings. A decision-maker cannot compare the direct-allocation benefit with the transfer-market alternative until the locked object is exact.

The wider regional constraint is real. A recent CircleID analysis describes how the /23 ceiling and uneven market availability can constrain larger deployments, and notes that prop-168 failed to reach consensus at APNIC 61. But a real demand problem does not settle the design. It makes queue transparency and mobility terms more important.

APNIC 62 needs three reproducible states

The September meeting does not need to choose between conservation and growth as slogans. It needs to make three state changes reproducible.

First, the pool state: on what date is the /12 reserved, which prefixes enter it, and what exact event closes ordinary distribution? Second, the queue state: when does an application obtain a position, what makes it complete, how are competing direct and NIR requests ordered, and what is published without exposing confidential business material? Third, the mobility state: which direct blocks become locked, which transfer types are covered, when the five years start, and what happens after a merger or account closure?

Lu Heng's Running-Code Primacy supplies the right discipline. A registry declaration should correspond to state that an operator can verify rather than an institutional promise that must be trusted. For prop-168, that does not require publishing private forecasts. It requires public pool totals, stable queue rules, dated decisions and a block-level transfer-eligibility result that two observers can reproduce.

The distributional cost of delay also belongs in the design. Heng Lu's Poverty Penalty argues that administrative proof and uncertainty fall hardest on networks with the least spare capital and staff time. A no-SLA queue can therefore reverse part of the proposed benefit: the applicant unable to bridge the wait buys the most expensive substitute.

No consensus result, final text, queue, demand count or implementation date existed at the research cutoff. The proposal remains scheduled for discussion at APNIC 62 on 10 September. The arithmetic is nevertheless already useful. A /22 ceiling describes how much an approved account might hold. The /12 subtraction, queue rule and five-year clock decide whether that possibility becomes usable infrastructure.

Sources