Summary

  • AFRINIC’s Phase 1 began on 31 March 2017 under a live /13 maximum. Draft 4, published on 14 April with the status Under Discussion, retained Draft 3’s proposed /18 maximum: 16,384 addresses rather than 524,288, making each maximum-sized request 32 times smaller.
  • The proposal combined that smaller ceiling with an eight-month planning horizon, a 90 per cent prior-use test, no explicit general recurrence limit, no explicit ordinary minimum, a /11 trigger into Phase 2 and a /22 Phase 2 maximum. Those controls shaped paperwork, timing and operational risk in different ways and must not be conflated.
  • Draft 4 removed Draft 3’s direct /16 critical-infrastructure and /14 new-entrant reserves, did not retain its /13 unforeseen-use reserve as such, and instead proposed a staff-evaluated /12 reserve for IPv6-deployment support, with individual access capped at /24.
  • A finite unallocated pool justifies a transparent private allocation procedure. It does not make AFRINIC a sovereign, legislature, regulator, police force, prosecutor, punishment authority, confiscator or court; its legitimate scarcity controls should be modelled, reasoned, reviewable and should end when an address leaves the free pool.

Summary

  • AFRINIC’s Phase 1 began on 31 March 2017 under a live /13 maximum. Draft 4, published on 14 April with the status Under Discussion, retained Draft 3’s proposed /18 maximum: 16,384 addresses rather than 524,288, making each maximum-sized request 32 times smaller.
  • The proposal combined that smaller ceiling with an eight-month planning horizon, a 90 per cent prior-use test, no explicit general recurrence limit, no explicit ordinary minimum, a /11 trigger into Phase 2 and a /22 Phase 2 maximum. Those controls shaped paperwork, timing and operational risk in different ways and must not be conflated.
  • Draft 4 removed Draft 3’s direct /16 critical-infrastructure and /14 new-entrant reserves, did not retain its /13 unforeseen-use reserve as such, and instead proposed a staff-evaluated /12 reserve for IPv6-deployment support, with individual access capped at /24.
  • A finite unallocated pool justifies a transparent private allocation procedure. It does not make AFRINIC a sovereign, legislature, regulator, police force, prosecutor, punishment authority, confiscator or court; its legitimate scarcity controls should be modelled, reasoned, reviewable and should end when an address leaves the free pool.

L3 — Fourteen days between two Phase 1 ceilings

The decisive interval began on 31 March 2017. AFRINIC’s dated notice says an approved request could no longer be fulfilled outside the organisation’s final /8, activating Phase 1 of the operative Soft Landing policy. The notice published on 3 April described the rules already in force: a maximum allocation or assignment of /13, a minimum of /22 for a Local Internet Registry and /24 for an End User, an eight-month planning window, and a requirement that 90 per cent of previously allocated or assigned resources be used before a further request. It did not state a general recurrence limit. It also identified a /12 held for unforeseen future uses. These details establish the live administrative position; they do not describe Draft 4.

On 14 April, exactly fourteen days after the trigger, AFRINIC published AFPUB-2016-V4-001-DRAFT-04. The document identified itself as version 4.0 and Under Discussion. It was a proposal by Omo Oaiya, Joe Kimaili and Alain P. Aina, not a rule already applied to applicants. The distinction is more than procedural housekeeping. Comparing a draft with an earlier draft reveals what its authors changed on paper. Comparing it with the live policy reveals what an applicant would have encountered if the proposal had displaced the rule that had just begun to operate.

Draft 4 matters here because those two comparisons lead to different, equally necessary conclusions.

Against the live policy, the most visible difference was the Phase 1 maximum. A /13 contains 524,288 IPv4 addresses. A /18 contains 16,384. Draft 4’s proposed maximum was therefore one thirty-second of the live maximum, or 32 times smaller. Prefix notation can disguise the scale: five additional bits in the prefix do not sound dramatic, yet each step halves the number of addresses in a block. Five halvings turn a request capable of covering 524,288 addresses into one capped at 16,384. That arithmetic defines the permitted size of a request event. It does not show the amount any applicant actually received, the number of applications that would have passed, or the duration of the free pool.

Against Draft 3, however, the /18 was not new. Draft 3 had already proposed the same Phase 1 maximum. The earlier movement from /15 to /18 belongs to the history of Draft 3, not to the Draft-3-to-Draft-4 redline. Draft 4 retained /18, just as it retained a /11 threshold for entering Phase 2 and a /22 Phase 2 maximum. Calling the /18 a Draft 4 term is correct. Saying Draft 4 invented that figure is not. The precise formulation is that Draft 4 carried the /18 forward from Draft 3 and, if adopted, would have replaced the then-live /13 ceiling with it.

The distinction between /11 and /22 is just as important. Under Draft 4, Phase 2 would begin when no more than a /11 of non-reserved space remained in the final /8. That /11 was an inventory trigger: it described the point at which the rule would change. Once Phase 2 began, /22 would be the maximum allocation or assignment per request. The first number governed the timing of a phase transition; the second governed the size of a request within that phase. Neither was evidence that a /11 or /22 had been issued, and the /11 was never a per-request allowance.

Draft 4 also retained the eight-month planning period. Applicants would justify needs expected over that horizon, rather than present an indefinitely expansive forecast. This could have tied applications to nearer-term evidence. But an eight-month needs window is not a command to wait eight months before returning. The draft stated no general recurrence interval for ordinary requests. It also retained the 90 per cent utilisation test for previously received resources, with an exception for a first request from a new Local Internet Registry or End User. The planning horizon, utilisation threshold and recurrence rule are separate levers.

Draft 4 specified the first two and left the third without an explicit general limit.

That separation matters when reconstructing the application path. An established applicant would have had to demonstrate use of 90 per cent of earlier resources and document needs within the planning window. A first-time applicant would not face the prior-use test because there was no prior allocation or assignment to measure. Any approved ordinary request would be bounded by the phase maximum.

Yet the absence of a stated general recurrence interval meant that the text did not say, in a single transparent rule, how quickly a successful recipient could return, how related requests would be aggregated, or how repeated applications would affect queue position. One cannot fill that silence with a restriction introduced in another version.

The minimum was another point of divergence. The operative Phase 1 notice stated /22 for Local Internet Registries and /24 for End Users. Draft 4’s body contained no explicit ordinary minimum. AFRINIC staff later observed that the wording could extend in principle to a /32 and raised practical concerns about routability below /24. That observation is evidence of how staff read the draft and of a drafting problem; it is not evidence that AFRINIC actually issued individual /32 allocations, nor that such tiny blocks would have been globally routable. At AFRINIC-26 the meeting record said the authors agreed, in response to comments, to retain /24 as a minimum. That anticipated a textual correction. It did not silently rewrite the 14 April document.

The later restoration of an ordinary /24 minimum and introduction of a 24-month recurrence restriction belonged to Draft 5, not Draft 4. Their relevance here is strictly boundary-setting: they cannot be read backwards into the selected text. Draft 4 must be assessed as it stood, with no explicit general recurrence limit and no explicit ordinary minimum. Otherwise safeguards written later would falsely improve the earlier proposal, concealing the very administrative ambiguities that its staff assessment and meeting discussion exposed.

The most substantial Draft-3-to-Draft-4 changes concerned the treatment of reserves. Draft 3 had set aside a direct /16 for critical Internet infrastructure and a direct /14 for new Local Internet Registries and End Users. It also carried a /13 reserve for unforeseen use. Draft 4 removed the two constituency-specific reserves and did not keep the unforeseen-use reserve in that form. In their place, it proposed a contiguous /12, containing 1,048,576 addresses, dedicated to supporting IPv6 deployment once AFRINIC could no longer satisfy ordinary requests. This was not ordinary Phase 1 inventory waiting to be claimed through the standard path.

Access to that proposed /12 was narrower than its headline size suggests. An applicant had to demonstrate a need for IPv4 resources supporting IPv6 deployment, show that no suitable alternative allocation was available, and have received nothing under the reserve policy during the preceding six months. An individual grant was capped at /24, or 256 addresses. Draft 4 gave AFRINIC staff a role in judging the justification and suggested sparse allocation within the reserved block. The six-month condition applied to prior receipt from this particular reserve. It was not a general six-month interval between ordinary applications and cannot be used to fill the draft’s wider recurrence gap.

Arithmetic supplies one useful illustration and one tempting error. A /12 can be divided arithmetically into 4,096 blocks of /24. That does not mean the reserve guaranteed 4,096 recipients. Eligibility, actual demand, fragmentation, sparse allocation, administration and the possibility of different operational outcomes all stand between capacity and recipients. The exact, defensible claim is that Draft 4 protected a large block from ordinary allocation and capped an eligible reserve award at /24. It thereby located option value in a special-purpose chamber; it did not specify the realised use of that option.

The staff assessment dated 15 May helps identify unresolved mechanics. Staff noted the absence of an explicit minimum, asked for clarity about the reserve replacement and its separation from ordinary space, and raised the operational burden created by repeated requests. It also examined the discretion staff would exercise over reserve justifications. Those comments are valuable because they show the institution itself recognising implementation questions. They are not independent findings by a regulator or court. The assessment’s statement that no legal-counsel comments were observed is an absence of recorded comment, not legal approval.

The procedural ending is equally limited. AFRINIC-26 recorded the draft’s /18, /22 and eight-month terms, captured discussion of the minimum and recorded no consensus, returning the proposal to the list. Draft 4 remained an under-discussion private proposal. Publication did not make it binding; a policy meeting did not turn it into law; a revision-history statement that changes reflected consensus did not prove region-wide assent. The official record establishes text, timing, institutional commentary and a no-consensus disposition. It does not establish adoption, implementation or application to a named member.

This careful reconstruction narrows the proper inquiry. Draft 4 was neither an accomplished redistribution nor an empty collection of prefix lengths. It was a proposed administrative design placed beside a live system only two weeks after activation. Its terms would have changed the maximum size of an ordinary Phase 1 request, the presence of ordinary minima and the allocation of reserves. Other terms—especially /18 itself, the /11 trigger, /22 in Phase 2, the eight-month horizon and the 90 per cent test—were continuations from Draft 3. The institutional question begins with knowing which parts moved, which parts stayed, and which live rules the whole package would have displaced.