Summary

  • APNIC lists prop-164 version 6 for discussion at the APNIC 62 Open Policy Meeting; it was not adopted or implemented at the reporting freeze.
  • The draft would lower the minimum IPv6 allocation to /40, align allocation sizes on four-bit boundaries and create explicit rules for reserving nearby space for future growth.
  • Better downstream registration is a credible objective, but the reserved space, return path and interaction with prop-170 need measurable public accounting.

An organisation can need less than a /32 yet still need to document the prefixes it delegates to customers or internal operating units. Prop-164 starts from that mismatch. Its authors say holders of longer prefixes cannot make those downstream allocations and maintain the corresponding Whois or RDAP records under the present framework without requesting a /32.

APNIC records version 6 as posted on 2 September 2026 and lists the proposal for discussion at the APNIC 62 Open Policy Meeting. The Secretariat impact assessment shown on the proposal page was prepared for version 5 and was posted on 18 August. These are stages in an open policy process, not evidence that the community has reached consensus or that APNIC has implemented the changes.

The current text would replace the /32 minimum IPv6 allocation with /40. It would also change the cited eligibility rule for an account holder with existing IPv4 space from a /32 IPv6 block to a /40. When an LIR delegates to a directly connected ISP, the LIR would have to place the relevant delegation details in Whois.

Those changes connect resource sizing to record quality. A smaller starting allocation could let an account holder receive only the space it needs while still recording downstream delegations as allocations. That is the proposal's strongest public-interest case: better registry records without requiring every qualifying network to take the current minimum.

Four bits at a time

Prop-164 does not propose arbitrary prefix sizes. It says initial and subsequent allocations must align to the next-shorter four-bit, or nibble, boundary. Its example turns a justified /42 request into a /40 allocation without requiring additional justification for the two-bit difference.

Nibble boundaries make IPv6 prefixes easier to express in hexadecimal, but they also mean the delegated block can be larger than the exact calculated requirement. Unlike prop-170, which retains a /32 minimum and addresses conditional alignment for larger LIR allocations, prop-164 changes the minimum itself and couples it to explicit reservation and resizing rules. The two proposals overlap, but they are not substitutes.

The proposal would allow an account holder with a /32 or shorter allocation to reduce it to /36 or /40 by keeping the first block of the chosen size and returning the balance. It would also let holders of allocations from /40 through /33 expand to /32 without new justification if they satisfy the referenced eligibility criteria.

These movements make reversibility central. A smaller initial allocation is useful only if later expansion is dependable; a return option is credible only if the registry can describe what was returned, what remains reserved and what happens when adjacent space is unavailable.

The reservation is the policy's real ledger

For requests longer than /28, version 6 says APNIC would make a sparse allocation from a /28 and reserve the difference for the same account holder's future allocations. If a suitable /28 is unavailable, APNIC would allocate the requested size from the largest contiguous space available and make no reservation. If APNIC exhausted its available IPv6 pool and could not secure more, reserved space would be released back to the free pool.

That structure separates three states that public reporting should never blur: delegated space, reserved-but-not-delegated space, and free space. Only the first is currently held by the account holder. The second is a conditional registry commitment. The third remains available to others.

The proposal's claimed benefit is more accurate registration, so the reservation mechanism should meet the same standard. APNIC need not expose confidential network plans to publish aggregate measures: how many allocations were below /32, how much space was reserved around them, how often holders expanded into a reservation, how often adjacency was unavailable, how many allocations were reduced, and how much reserved space returned to the free pool.

Without those measures, smaller allocations could improve the visible Whois record while moving a material part of resource management into an invisible reservation ledger. With them, the community could test whether the policy conserves delegated space, preserves practical growth paths and treats comparable applicants consistently.

Version boundaries matter

APNIC's displayed impact assessment applies to version 5. It says systems and procedures would need to support allocations smaller than /32, returns, resizing, nibble-aligned sizes and explicit reservation handling. It also raises questions about what happens when a suitable /28 is unavailable and whether explicit reservations are necessary alongside APNIC's existing sparse delegation practice.

The assessment asked whether holders of a /32 or larger block could reduce to /36 or /40, or only /40 under the text it reviewed. Version 6 now expressly names both /36 and /40. That appears to answer the narrow drafting question, but APNIC has not labelled the version 5 assessment as a fresh assessment of version 6. Readers should therefore avoid treating every operational conclusion as revalidated for the new text.

The Secretariat identified no material legal issue and estimated roughly nine months for implementation if consensus were reached, subject to editorial comments. It also said prop-164 and prop-170 would require drafting revisions if both reached consensus. The final public text, not internal guidance, should resolve which sizing and alignment rule governs each request.

The proposers list no known disadvantages and no known impact on resource holders. Those statements describe their assessment. They do not establish that registry changes, applicant choices, NIR consistency, returned-space handling or reserved inventory carry no cost.

What remains unknown

It is not known whether the community will reach consensus or amend the proposal, how many account holders would request /36 or /40, how much space would be reserved or for how long, whether APNIC and NIR decisions would remain consistent, or how the final text would resolve its overlap with prop-170. These are unresolved questions, not established effects of the proposal.

Sources