Summary

  • Prop-164-v005 would lower APNIC's minimum IPv6 allocation from /32 to /40, introduce smaller nibble-aligned allocations, and add return, reduction and reservation machinery.
  • Prop-170-v001 would keep the /32 minimum and conditionally align an allocation only after an applicant requests it and the existing needs test validates the underlying requirement.
  • The Secretariat's 18 August assessments say the texts would need revision if both reach consensus. Before APNIC 62 judges them, the community needs to see what 164 only, 170 only, both and neither actually authorize.

One label now controls two different baselines

The NOG Alliance policy overview lists prop-164 and prop-170 separately for discussion at APNIC 62. That separation matters more than the nearly identical language in their titles.

Prop-164 changes who can receive what at the bottom of the allocation scale. Its current fifth version would replace the /32 floor with /40, allow nibble-aligned steps such as /32, /36 and /40, and introduce a route for some existing holders to return excess space and reduce their allocation. It also writes sparse placement and nearby reservation into the design so a smaller starting block can grow later.

Prop-170-v001 leaves that floor alone. It applies to a request that has already passed the ordinary needs assessment. If the applicant explicitly asks for alignment, the resulting allocation may be rounded to a four-bit boundary. It does not apply to assignments and does not generally order existing holders to return, resize or renumber.

Those are not two wordings for the same result. One expands the set of permissible small allocations and creates a return-and-reservation path. The other adds a conditional sizing step to the existing allocation process. A community member can support one and oppose the other without contradicting a view about hexadecimal readability.

The history makes that distinction easier to see. An external archive of prop-164's earlier text framed the problem as a choice between taking an unnecessarily large allocation and accepting an assignment that could not represent downstream sub-allocations as intended. A JPNIC review explained the same allocation-versus-assignment dilemma for Japanese readers. The proposal has since moved from /36 to /40 and acquired more explicit growth machinery. Its decision surface has grown, not merely its version number.

The 18 August warning is not an editorial footnote

The current impact assessments, posted on 18 August, attach different implementation work to each proposal. Prop-164's assessment estimates roughly nine months and would require registry support for smaller allocations, reduction requests and reservation handling. It asks whether an existing holder may reduce to /36 or only /40, what happens if suitable adjacent space is unavailable, and whether explicit reservations add anything to existing sparse placement.

Prop-170's assessment estimates roughly six months. It needs guidance and assessment changes for conditional alignment while preserving the needs-based framework. The assessment says the /32 minimum remains, assignments remain outside scope and existing holders normally do not have to move.

Then both records add the same warning: “If prop-164 and prop-170 both reached consensus, drafting revisions would be required to ensure the policies do not conflict.”

That sentence creates a decision problem. Editorial work can reconcile numbering, terminology and cross-references after a community judgment. It should not decide whether /32 remains the minimum, whether a holder has a new return path, whether nearby space must be reserved, or whether an applicant enters one assessment route rather than another. Those are policy states with different beneficiaries, costs and operational consequences.

Earlier operator discussion already showed why a generic vote on smaller blocks would provide too little evidence. Feedback preserved by Mail Archive asked whether moving from /32 to /36 would simply revive pressure for /40 later and challenged the claimed conservation benefit. A SANOG policy update documented the registry-accuracy case for the earlier version. Those positions concern the minimum itself. They cannot be treated as automatic evidence about prop-170's conditional alignment after a need has already been validated.

Four outcomes need four visible rows

The repair is a short public matrix before the 10 September Open Policy Meeting.

Under prop-164 only, the matrix should state the new minimum, permitted nibble steps, treatment of existing holders, return conditions, reservation rule, adjacency fallback and implementation period. Prop-170's conditional alignment mechanism would not silently appear.

Under prop-170 only, /32 would remain the minimum. The matrix should identify the request, validation and alignment stages, confirm that assignments remain outside scope, and state what happens when contiguous expansion is impossible. Prop-164's /40 entry path and return mechanism would not follow.

Under both, the community needs the actual combined rule. Which proposal controls overlapping definitions? Does the /40 minimum coexist with a separate post-assessment alignment step? Can an existing holder reduce to /36, /40, or either? Is nearby capacity reserved by rule, by operating practice, or not at all? Which provisions govern an expansion when the reserved span is unavailable? A note promising later harmonization is not an answer.

Under neither, the present baseline remains. That outcome should also be explicit, because no consensus is not permission to implement the shared idea through guidance.

The matrix does not require the chairs to predict the room. It makes the question stable enough for the room to answer. If no agreed combined text exists before APNIC 62, the two consensus judgments can remain separate and any material reconciliation can return for visible review.

Prefix aesthetics do not settle institutional authority

Nibble boundaries are useful. In hexadecimal notation, every digit represents four bits, so four-bit boundaries make prefixes easier to read, divide, document and audit. RFC 6177 also rejects the idea that one end-site size fits every network and supports more considered sizing decisions.

But a useful technical property does not answer every governance question attached to it. APNIC's community process controls whether a proposal reaches consensus. The chairs control the formal determination. Editors can prepare the resulting text. The Secretariat controls implementation in registry systems and procedures. Resource holders control how an allocation is subnetted, documented, routed and used. Whois and RDAP record the delegated state; they do not prove that every reserved address is deployed or that every operational plan succeeded.

Heng Lu's Minimum Initial Specification supplies a useful discipline here: put only the common rule required for coordination in the shared layer, and leave later operational choices attributable to those who make them. That does not prohibit an allocation policy. It requires the policy to say exactly which shared state it changes and which choices remain with running networks.

The two proposals can therefore be judged on evidence rather than vocabulary. Does a /40 minimum improve registry accuracy enough to justify new return and reservation machinery? Does conditional rounding preserve need while producing more usable address plans? Can each mechanism be implemented without turning an internal placement assumption into an entitlement? These are separate questions.

The meeting should decide the text it can name

Nothing reviewed for this article establishes that either proposal will pass. A sixth version, an amendment or a combined draft may arrive before 10 September. The implementation estimates may change. No real applicant outcome or routing effect can be inferred from policy text that is still under discussion.

The verified point is narrower. APNIC's own current assessments identify a possible conflict but defer its resolution until after the condition that would activate it: consensus on both proposals. That sequence can be improved before the meeting.

Publish the four rows. Attach every material clause to the proposal or combined text that authorizes it. Ask for separate judgments where the consequences differ. If both receive support but the combination introduces a new minimum, return right, reservation duty or scope rule that community members have not seen, publish the combined version and review it as a proposition, not as punctuation.

APNIC 62 need not choose between tidy hexadecimal plans and accountable process. It can have both. The first step is to stop allowing one word—nibble—to stand in for two policies.

Sources