Summary

  • Draft 6 answered three identifiable drafting problems, but the official record supplies a change log rather than a complete redline or proof that every concern was closed.
  • Clear distinctions between holdings and receipts, a calculable 24-month recurrence limit and a single replacement /12 reduce administrative uncertainty while still distributing delay, opportunity and cost among operators.
  • A legitimate private-bookkeeper response is an objection-and-calculation ledger, transparent queue arithmetic, reasoned reserve administration, market and portability alternatives, and a firm sunset on allocation-era control once addresses leave the free pool.

L3 — Three edits before the room could say yes

The significance of Draft 6 lies in its smallness. AFRINIC’s official history places it sixth in a seven-version chain and dates it to 22 September 2017. Its recorded change log contains three items, no more: section 5.4.3 was clarified so that demonstrated need applied across the exhaustion phases and aligned with other provisions; section 5.4.6 was reworded for clarity; and section 5.4.7 was reworded to make clear that the proposed IPv6-deployment /12 replaced the previous /12 reserve. That list is enough to identify the institutional act. It is not enough to manufacture a full textual comparison, infer unrecorded concessions or pronounce every prior concern resolved.

The baseline itself needs careful dating. AFRINIC’s proposal history and archive index record Draft 5 on 27 June 2017, while the Draft 5 detail page displays a submitted date of 29 June and labels the proposal Under Discussion. The closed record does not reconcile those two June dates, so neither should an account of the September revision. What matters for the delta is that staff assessed Draft 5 on 24 August and identified operational ambiguities before Draft 6 appeared. Preserving the discrepancy is not pedantry.

A chronology that openly carries an unresolved two-day difference demonstrates the same discipline that the policy text required: when the record cannot settle a point, the account should not pretend that it can.

Draft 5’s allocation geometry gave those ambiguities practical weight. It used a Phase 1 maximum of /18 and a Phase 2 maximum of /22, with /24 as the minimum in each phase. In address counts, those prefixes represented 16,384, 1,024 and 256 addresses respectively. The proposal also used an eight-month planning window and a 90 per cent prior-use test, subject to a first-request exception. These elements formed a rule for rationing a dwindling administrative inventory. They were not just labels. They set the quantities an applicant might seek, the evidence it might have to assemble and the assumptions through which staff would decide whether a request entered or remained outside the queue.

The first September repair concerned demonstrated need. Draft 5 section 5.4.3 already said that allocations and assignments in either exhaustion phase would be based on justified and demonstrated need. Staff nevertheless saw a need to clarify how that standard related across phases and to the surrounding policy framework. The official Draft 6 summary says that section 5.4.3 was clarified for all phases and aligned with other CPM sections. That is a meaningful drafting gain: the decision-maker and applicant are less likely to treat need as disappearing, changing character without explanation or applying only on one side of a phase boundary.

It remains an official description of the revision, not evidence that a particular need test was efficient, fair or independently authorised.

The second repair addressed the most consequential ambiguity in the staff assessment. Draft 5 section 5.4.6.1 referred to an organisation’s total allocations or assignments not exceeding /18 in Phase 1 or /22 in Phase 2. Read literally, “total” could capture everything the organisation already held, not merely what it received under the exhaustion phase. That distinction can separate a longstanding operator with substantial deployed holdings from a newcomer even where both seek the same small increment from the remaining pool. Staff explicitly identified the risk that the clause described total holdings rather than additional receipts. The uncertainty was therefore not invented by a later critic; it was an implementation problem recorded within the official process.

Draft 5 section 5.4.6.2 added a temporal complication. It allowed another round after a 24-calendar-month wait once an organisation had received the maximum, but staff described the override as difficult to understand. Staff suggested framing the rule around one or more allocations or assignments totalling /18 within any 24-month period in Phase 1 and /22 within any 24-month period in Phase 2. That suggestion has the virtue of turning an uncertain reset into an arithmetic test. Even so, the evidentiary boundary is strict: the official Draft 6 log says section 5.4.6 was reworded for better clarity, but the available record does not reproduce a verified full redline or establish that staff’s suggested words were adopted verbatim.

That boundary changes how the September act should be described. It is safe to say that the recurrence provision was reworded in response to a field of ambiguity that included holdings, additional receipts and a difficult 24-month override. It is not safe to quote an imagined Draft 6 formula or to say that every edge case disappeared. “Reworded for clarity” identifies an intention and a revision category. It does not reveal every operand, timestamp convention or treatment of multiple receipts. A reader can credit the direction of travel while withholding a conclusion about the exact destination.

The distinction is especially important when a rule will be executed by staff against applications whose timing and history differ.

The third repair concerned a block much larger than the ordinary phase maxima. Draft 5 proposed a contiguous /12 from the final /8 to support IPv6 deployment after ordinary available space could no longer meet requests. A /12 contains 1,048,576 IPv4 addresses. Staff asked whether this new reserve coexisted with an earlier /12 reserved for unforeseen future uses and proposed explicit one-for-one replacement wording. Without that clarification, the draft could be read as sequestering two separate million-address blocks. The official Draft 6 log closes that particular binary question by stating that the proposal’s reserve replaced the previous reserved /12.

Replacement clarity is a substantial improvement, but its proper scope is narrow. It answers “one or two?” at the level recorded in the official summary. It does not, by itself, answer every question about reserve eligibility, the minimum available from the reserve, requests involving IDN ccTLDs or the wording of a technology example. Nor does it show who would eventually receive space, whether the reserve improved IPv6 deployment or how much demand it faced. One should neither trivialise the removal of a possible duplicate reserve nor promote it into proof of complete policy settlement.

The gain is coherent inventory geometry: one stated replacement block rather than an unintended doubling of withheld stock.

Taken together, the three edits present the strongest benign case for Draft 6. Mature drafting is supposed to narrow uncertainty discovered by implementers. Applying demonstrated need consistently across phases can reduce divergent evaluations. A clearer recurrence rule can make limits calculable instead of leaving applicants to guess how prior holdings and later receipts interact. An explicit replacement reserve can prevent duplicate sequestration. Better text reduces interpretation costs, allows staff to apply a more stable frame and gives participants a more definite instrument to examine.

Nothing in a critical institutional analysis requires denying those benefits.

But the benign case cannot bear more weight than the record gives it. Publication on 22 September did not establish adoption, ratification, implementation, application, consensus or last-call status. The official change log does not say that all staff comments, mailing-list objections or floor concerns had been answered. Continued drafting proves that text changed; it does not prove that absent operators consented or that a meeting process represented every network exposed to the rule’s economic effects. Clearer language can make a private queue more reviewable.

It cannot turn the queue administrator into a public sovereign or convert procedural movement into authority.

The later chronology supplies a useful boundary rather than retrospective validation. AFRINIC-27 presented Version 6 on 30 November 2017, recorded further proposed wording changes and included a co-chair declaration of consensus that would send the proposal onwards. AFRINIC’s version summary then records Draft 7 on 1 December as the last-call version. Those later events show why the September publication must remain a pre-last-call instrument with text still open to alteration. They do not permit the later declaration or later wording to be backdated into Draft 6, and they do not prove what the September text achieved in practice.

That sequence also exposes the difference between a response record and an objection ledger. A change log says what the drafter reports changing. A response record may connect a staff concern to a revision. An objection ledger goes further: it identifies each competing reading, states the adopted disposition, shows the resulting calculation, records what remained unresolved and gives reasons. Draft 6’s three recorded items provide enough material to reconstruct the first two layers cautiously. They do not provide the last.

Treating the appearance of successive drafts as evidence of closure would erase precisely the questions that a late-stage instrument should make easiest to inspect.

Official records retain an indispensable but bounded role in this reconstruction. AFRINIC’s pages prove proposal identity, dates, staff concerns, the three-item summary and the later procedural boundary. Staff observations illuminate how a clause might have been administered. Neither source type establishes a public-law mandate merely by describing a “community” process or a policy objective. An institution can give authoritative evidence about what it published without becoming the final authority on the legitimacy or economic wisdom of what it proposed.

That division of evidentiary labour is not hostility to technical coordination; it is what keeps coordination accountable to its actual function.

NRS’s account of AFRINIC’s private membership and allocation context places the organisation on the correct institutional footing. Heng Lu’s analysis supplies the controlling boundary: a registry record describes resource coordination but does not create the underlying reality or acquire sovereign title to it. LARUS makes the operator consequence legible, especially for smaller businesses facing constrained supply and considering leasing as an alternative. BTW connects eligibility, timing and reserve decisions to capital value, operational optionality and continuity. These are not decorative counterweights to an official narrative.

Together they explain why exact drafting matters and why exact drafting still cannot enlarge a private bookkeeper’s authority.

AFRINIC’s legitimate task here was both useful and limited. As a private technical bookkeeper and coordinator, it could preserve uniqueness, maintain accurate records, verify facts material to an objective queue and administer genuinely unallocated stock. It could set a minimum, define phase maxima, use a documented utilisation test and planning window, calculate recurrence and reserve some of its unallocated inventory for a technically stated purpose. None of those activities made it a government, legislature, regulator, police force, prosecutor, punishment authority, confiscator or court.

The limits are part of the mechanism, not a rhetorical aside.

The September record therefore supports a measured conclusion. Draft 6 narrowed three real ambiguities before a later meeting could consider the proposal. Its edits improved the prospects for consistent administration, particularly by aligning demonstrated need, revisiting recurrence and eliminating a two-reserve reading. Yet the publication did not itself disclose the exact full recurrence calculation, model effects across applicant cohorts or prove that residual objections were exhausted. The right description is a better specified draft for a private queue.

Calling it less would miss the value of the repairs; calling it consensus or last call would rewrite the chronology.