Summary

  • Draft 5 made one direct, documented floor-to-text repair by restoring /24 as the minimum in both phases, while leaving the /18 and /22 maxima in place.
  • It did not adopt the suggestion that the /12 IPv6-deployment reserve should be confined to new LIRs; its actual gates concerned need, a six-month interval and the absence of another suitable allocation.
  • The absence of a Board disposal role was already recognised as a Draft 4 feature, while the authors rejected the asserted withdrawal premise and continued the proposal.
  • The new 24-month recurrence rule resembled Soft Landing SD’s two-year design, but resemblance, acknowledgement and a call for convergence do not prove who caused the clause.
  • The defensible institutional model is a thin, visible queue for unallocated resources—not sovereign regulation—with a public disposition ledger, objective calculations, reasoned review and practical market or portability alternatives.

Summary

  • Draft 5 made one direct, documented floor-to-text repair by restoring /24 as the minimum in both phases, while leaving the /18 and /22 maxima in place.
  • It did not adopt the suggestion that the /12 IPv6-deployment reserve should be confined to new LIRs; its actual gates concerned need, a six-month interval and the absence of another suitable allocation.
  • The absence of a Board disposal role was already recognised as a Draft 4 feature, while the authors rejected the asserted withdrawal premise and continued the proposal.
  • The new 24-month recurrence rule resembled Soft Landing SD’s two-year design, but resemblance, acknowledgement and a call for convergence do not prove who caused the clause.
  • The defensible institutional model is a thin, visible queue for unallocated resources—not sovereign regulation—with a public disposition ledger, objective calculations, reasoned review and practical market or portability alternatives.

Summary

  • Draft 5 made one direct, documented floor-to-text repair by restoring /24 as the minimum in both phases, while leaving the /18 and /22 maxima in place.
  • It did not adopt the suggestion that the /12 IPv6-deployment reserve should be confined to new LIRs; its actual gates concerned need, a six-month interval and the absence of another suitable allocation.
  • The absence of a Board disposal role was already recognised as a Draft 4 feature, while the authors rejected the asserted withdrawal premise and continued the proposal.
  • The new 24-month recurrence rule resembled Soft Landing SD’s two-year design, but resemblance, acknowledgement and a call for convergence do not prove who caused the clause.
  • The defensible institutional model is a thin, visible queue for unallocated resources—not sovereign regulation—with a public disposition ledger, objective calculations, reasoned review and practical market or portability alternatives.

L3 — What the floor changed, and what Draft 5 left alone

The meeting record begins with a draft, not a law. Draft 4 of AFPUB-2016-V4-001 was presented on 31 May with a /18 maximum in the first exhaustion phase, a /22 maximum in the second, an eight-month planning horizon and a dedicated reserve intended to assist IPv6 deployment. The co-chairs recorded no consensus. That outcome is the first guardrail around the history. It cannot be restated as approval, and it cannot be used as a shortcut to adoption, ratification, implementation or last call. Draft 5, when it appeared in June, was still an under-discussion proposal.

The surviving record does not establish that any operator ever received or was denied space under its terms.

The second guardrail is chronological modesty. AFRINIC’s revision history and archive list Draft 5 on 27 June 2017. The Draft 5 detail page displays a submission date of 29 June. Nothing in the available record explains whether one date marks completion, posting, receipt or some other administrative step. The two dates should therefore sit beside each other. Compressing them into a supposedly definitive publication date would improve the narrative at the expense of the evidence.

The disagreement is small in calendar time but important in method: an institution that keeps records should not be granted a reconciliation its records do not supply.

Against those limits, one response is unusually clear. Before the meeting, AFRINIC staff had identified the absence of an explicit minimum in Draft 4. Taken literally, that silence implied allocations as small as /32. Staff said such fragments would cause operational difficulty and require software changes, and recommended /24 instead. At AFRINIC-26 the authors expressly agreed to retain a /24 minimum. Draft 5 then specified /24 as the minimum in both phases. Here the chain has each needed link: an identified implementation problem, an author response in the meeting record and a matching provision in the next draft.

It is the strongest instance of demonstrated responsiveness.

That change did not alter the top of either phase. Draft 5 kept /18 as the Phase 1 maximum and /22 as the Phase 2 maximum. The proposal thus defined a corridor rather than a single allocation size. In the first phase, a request could be bounded below by /24 and above by /18; in the second, it could be bounded below by /24 and above by /22. The numerical differences make the architecture concrete. A /18 contains 16,384 IPv4 addresses, a /22 contains 1,024 and a /24 contains 256.

Moving from the first-phase maximum to the second-phase maximum reduced the largest one-round quantity by a factor of sixteen, while the restored floor avoided descending into tiny fragments whose administration staff had found impractical.

The proper description of the /24 result is precise: incorporated. It should not be inflated into a claim that Draft 5 accepted the meeting as a whole, nor diminished into a vague statement that authors merely “considered feedback”. The clause changed in the direction announced. That is what the record proves. A disposition account ought to preserve that granularity because institutional legitimacy is often lost in verbs. “Discussed”, “addressed” and “resolved” are not synonyms. An objection can be heard but rejected, answered outside the text, carried forward from an earlier draft, or incorporated word for word.

Only the last of those fits the /24 minimum.

The suggestion concerning the /12 reserve has a different disposition. The minutes record a proposal that access to the IPv6-deployment reserve should be restricted to new Local Internet Registries. Draft 5 did not create that class restriction. Its stated reserve gates instead concerned a demonstrated need connected to IPv6 support, a six-month interval since a previous receipt and whether another allocation could satisfy the need. It also contained a provision for Core DNS service providers. None of those conditions is equivalent to saying “new LIRs only”.

A need test asks what function the requested space will serve; an interval test asks when the applicant last received space; an alternatives test asks whether another allocation can meet the requirement. A status rule confined to new LIRs would ask who the applicant is. Draft 5 chose the former structure, not the latter.

There is a nearby provision that could confuse the account if read too quickly. For the ordinary prior-utilisation requirement, Draft 5 exempted a new LIR or End User making a first request from proving 90% use of earlier allocations or assignments. But that first-request exception was not the floor’s requested reserve limitation, and it was already present in Draft 4. It cannot be repurposed as evidence that the new-LIR-only reserve suggestion was accepted.

The two rules operate on different questions: one concerns proof of prior use where no prior allocation may exist; the other would have defined eligibility for a large, separately gated reserve. Treating them as interchangeable would manufacture a response that the text does not contain.

The reserve suggestion should therefore be marked not incorporated. That does not establish neglect, hostility or bad faith. The closed record does not contain a full author-by-author explanation, and minutes cannot reproduce every intervention or online comment. It establishes only that the meeting recorded the suggestion and that the next draft did not embody it. This is exactly why a response ledger would have been valuable. A short entry could have stated whether the authors rejected the new-LIR-only class, thought the need-based gates already answered the concern, or deferred the question.

Without that entry, readers can compare texts, but they cannot safely supply the missing reason.

The Board point requires a third label: carried forward. The floor record includes an observation that the Board should not determine the use of the reserved block. Yet the same meeting account described Draft 4 as having addressed that concern. Draft 5 likewise contained no Board disposal role. It would therefore be false to narrate the later draft as removing a power that Draft 4 still conferred. The better account is that the meeting confirmed or endorsed an existing boundary and Draft 5 retained it. The floor comment matters, but it matters as recognition of continuity, not as the cause of a fresh deletion.

This distinction protects both credit and criticism from exaggeration. If every desirable feature in a later document is described as a new response, the policy process appears more adaptive than the redline supports. If every unchanged feature is ignored, the floor’s role in testing and affirming boundaries disappears. “Carried forward” avoids both mistakes. It records that participants raised the issue, that the earlier draft already lacked the contested role, and that Draft 5 did not reverse course. It does not transform the absence of a Board clause into proof that the meeting conferred or withdrew authority.

A private association’s internal draft cannot manufacture public power in either direction.

The withdrawal dispute calls for a fourth category. The minutes record an allegation that the authors had previously promised to withdraw the proposal. They denied that premise. Draft 5 was subsequently published, so the proposal continued rather than being withdrawn. Those facts do not allow an adjudication of who remembered the earlier exchange correctly. The meeting minutes prove competing statements, not the underlying promise. The response map can nonetheless state what happened procedurally: the authors rejected the asserted premise, and their subsequent conduct was consistent with that rejection.

Here, “rejected as a premise” is more accurate than either “ignored” or “resolved”. Withdrawal would have ended the proposal. Continuation shows that the authors did not accept the demand’s factual foundation or requested consequence. But no court-like finding emerged from the room, and none should be invented. AFRINIC was not adjudicating testimony. The co-chairs’ no-consensus decision did not decide the truth of the alleged promise. It described the proposal’s status in that session. Separating those questions is not pedantry; it prevents a procedural chairing decision from acquiring the false weight of a legal judgment.

The most tempting line to overstate concerns the 24-month recurrence rule. The Soft Landing SD session that followed the Draft 4 discussion described a /16 maximum in its first phase and a two-year bar on another allocation after an organisation consumed that maximum. The meeting record also contains a request that the two proposal teams combine their strongest points. Draft 5 then introduced clauses requiring an organisation that had received the maximum allowable prefix in a phase to wait 24 calendar months before another round in that phase. Draft 5 acknowledged the Soft Landing SD authors and the community.

The resemblance is substantial enough to matter.

It is not substantial enough to prove direct copying or speaker-specific causation. The record contains no author statement saying that a named intervention produced the clause, no drafting note assigning parentage and no complete account of other discussions the authors may have considered. A two-year wait and a 24-month wait describe the same duration, and the call for convergence supplies relevant context, but sequence plus similarity does not close the causal gap. The defensible disposition is therefore “partly reflected, causation unproved”.

Draft 5 moved towards a mechanism heard in the neighbouring proposal, while the exact route by which that mechanism entered the text remains uncertain.

That uncertainty also matters because Draft 5’s recurrence mechanism was not identical to every feature of Soft Landing SD. The recorded SD presentation paired its two-year rule with a /16 Phase 1 maximum. Draft 5 retained a /18 Phase 1 maximum. A response map should trace the element that converged without pretending the designs merged wholesale. The participants’ request to combine “best points” was itself a preference, not a specification. Draft 5’s acknowledgement is evidence of intellectual and procedural proximity; it is not a deed assigning ownership of each sentence.

Staff’s earlier assessment supplies another layer of the map. Draft 4 had no explicit recurrence limit during exhaustion. Staff had sought clearer definitions and phase rules, clearer need language, rate limiting, reserve clarification and better sectioning, as well as the /24 minimum. Draft 5’s issue-and-action table said that most Draft 4 staff issues had been addressed. It did make definition and structural repairs, restored the operational floor and added the recurrence provisions. Yet some entries were marked not applicable or left to staff rather than settled in policy text.

“Most” is therefore the official summary, not a licence to say “all”.

Even within the new recurrence language, later staff assessment recorded ambiguity over whether the relevant total meant an organisation’s total holdings or additional space received during a phase. That observation is useful without importing any later draft: it shows that adding a control does not automatically make the control fully determinate. A 24-month clock needs a clearly defined trigger. Does it start only when the new grant equals the phase maximum, or when holdings reach some aggregate level? The available wording and assessment do not justify pretending the drafting question had disappeared.

An auditable map of the meeting to Draft 5 therefore has five distinct entries. The /24 floor was incorporated. The new-LIR-only reserve suggestion was not incorporated. The Board’s non-role was carried forward. The withdrawal demand’s premise was rejected, without an adjudication of truth. The recurrence idea was partly reflected in a form resembling Soft Landing SD, while personal causation remains unproved. Around those floor responses sat a wider set of staff-driven repairs, many addressed and some left outside the text or to later administration.

This classification explains why the no-consensus outcome and the later draft are compatible. No consensus did not freeze the document. Authors could revise and continue an under-discussion proposal. Conversely, publication of a revised draft did not reverse the meeting result or transform it into agreement. A draft can respond after a failed consensus call without becoming operative. The meeting and the later text are different institutional acts: one recorded the absence of agreement in a session; the other placed another proposal version into discussion.

Nor did the room represent a sovereign population. Attendance at a Public Policy Meeting may provide useful technical knowledge and a place to test language. It does not make attendees the owners of all number resources or confer a legislative mandate over a service region. “Community”, “fairness” and “regional need” describe how participants and AFRINIC framed the problem. They are evidence of claims and preferences, not independent sources of jurisdiction. The legitimate force of the /24 repair comes from its operational reason and its place in a bounded free-pool queue, not from a fiction that the floor was a parliament.

Seen that way, Draft 5’s uneven responsiveness is not an anomaly to conceal. It is the normal substance of policy drafting: some comments alter text, some fail, some confirm what is already present, and some coincide with a change whose origin cannot be proved. The defect lies in leaving readers to reconstruct those categories from a meeting summary, two drafts, staff tables and archive dates. Where the consequences concern valuable operating capacity, the disposition itself should have been published as clearly as the clauses.