Summary

  • Published to the RPD mailing list on 9 February 2016, AFPUB-2016-V4-001-DRAFT-01 proposed reducing the Phase 1 maximum from the /13 that its authors attributed to the existing Soft Landing policy to /15. It would move to Phase 2 when no more than a /11 of non-reserved Final /8 space remained.
  • The draft combined that smaller ceiling with a 90 per cent utilisation test, an IPv6-resource prerequisite and named channels for critical infrastructure, first-time LIRs or End-Users, and unforeseen future use. These were meaningful choices about practical access to a scarce administrative pool, not mere technical housekeeping.
  • Its arithmetic was only partly reproducible. D1 placed a /16 infrastructure reserve and a /14 new-entrant reserve in the final /11, but described a /13 contingency reserve as coming from the Final /8. It supplied neither exact prefixes nor a map proving that every reserve was disjoint, so no unique general-pool remainder can responsibly be inferred.
  • AFRINIC could legitimately maintain a unique allocation ledger and apply a published request rule within a private administrative and contractual relationship. Publication, scarcity and consultation could not give it ownership of Internet addresses, sovereignty over Africa, regulatory jurisdiction, or police, punitive, confiscatory or adjudicative power.

A smaller gate into the last pool

The bargain began with three pieces of architecture. First, the largest Phase 1 allocation or assignment would fall from the /13 ceiling that the authors said applied under the existing Soft Landing policy to /15. Secondly, Phase 1 would end when the remaining non-reserved space in the Final /8 reached no more than a /11. Thirdly, not every address beyond that gate would be exposed to the same kind of claim: the proposal named separate reserves for critical Internet infrastructure, applicants with no previous AFRINIC IPv4 space and unforeseen future uses.

Those choices matter because a prefix ceiling is a distribution rule. A /13 contains 524,288 IPv4 addresses, while a /15 contains 131,072. On address capacity alone, therefore, the proposed maximum was one quarter of the maximum D1 attributed to the old policy. That comparison says nothing about how many organisations would apply, how much they would receive, how quickly they would use it or how long the pool would last. It does show what the proposal changed at the level of a single decision: no qualifying Phase 1 request could be met with an award as large as the previous stated ceiling.

This made the proposal a reset rather than a cosmetic amendment. Smaller maximum awards can keep one request from consuming as much of a dwindling pool and can distribute scarcity across more decision moments. Yet the same rule can require a growing network to return more often, submitting evidence anew and becoming more dependent on the consistency, speed and transparency of registry administration. The ceiling therefore allocated not only addresses but also transaction costs and exposure to decision-making. A fair account has to hold both effects in view.

The threshold was equally consequential. A /11 contains 2,097,152 addresses. D1 did not say that Phase 2 began when the Final /8 as a whole reached that size. Its proposed wording focused on the point at which no more than a /11 of non-reserved Final /8 space remained. Reserves were already part of the definition of the gate. To reproduce the transition, an operator would need to know which blocks counted as reserved, which counted as non-reserved, and what the ledger showed at a definite time. The prefix length alone was not enough.

Nor did the draft confine Phase 1 awards to a single visibly named block. It proposed that allocations or assignments could be made from the Final /8 or other IPv4 space available to AFRINIC until the threshold was reached. That makes a consolidated pool ledger essential. If more than one inventory source could satisfy requests, the administrator had to expose which inventory had changed and how the non-reserved Final /8 balance had been calculated. Otherwise two observers could agree on the words /11 and still disagree about whether the gate had been crossed.

The proposal also addressed the boundary case. Applications still in process when the Phase 2 threshold was reached would be evaluated under the new policy. That sentence could affect an applicant without changing the date on which it filed: the relevant rule could turn on the state of the pool while the request remained pending. D1 identifies no such applicant and proves no actual application of this provision. As a design choice, however, it makes the transition clock part of the substantive bargain. A reproducible system would have to record the application time, the threshold time, the status of pending work and the rule version applied.

One draft, two documentary dates

The documentary chronology is short but precise. The D1 detail panel gives 4 February 2016 as the submission date. The revision history on the draft and AFRINIC's archived proposal records give 9 February 2016 as the date Version 1 was posted to the RPD mailing list. These dates describe different acts. Submission can precede public posting, and the record does not establish the exact time or channel of the earlier act.

The institutional event considered here is the 9 February publication of AFPUB-2016-V4-001-DRAFT-01. The official page identified it as Version 1.0 and “Under Discussion”, and said that it would amend AFPUB-2010-v4-005 while obsoleting none. Omo Oaiya of WACREN, Joe Kimaili of Ubuntunet Alliance and Alain P. Aina of TRS were named as authors. The published material does not allocate separate arguments or motives among them, so their individual rationales should not be invented.

“Under Discussion” is an important limit. It tells us that D1 was a proposal placed before a discussion process, not a record of an operative award system. The surviving material supports the act of publication and the content of the proposal. It does not establish later adoption. It cannot tell us that any applicant received, lost, delayed or was refused an address block under these clauses. Analysis must therefore test the proposed architecture in its own terms rather than smuggling later outcomes back into the first draft.

The baseline requires the same discipline. D1 said that AFRINIC's Board had ratified the existing Soft Landing policy on 11 November 2011. It characterised that policy as having two exhaustion phases: a Phase 1 maximum of /13; a Phase 2 maximum of /22 and minimum of /24; no distinction between existing and new LIRs or End-Users; and no IPv6-deployment condition. That is the authors' account of the bargain they sought to change. It is useful because the draft framed each intervention against it, but it is not a substitute for an independent reconstruction or full redline of the 2011 instrument.

Against that stated baseline, D1 presented four interventions. It lowered the Phase 1 maximum. It required an IPv4 applicant during exhaustion to possess IPv6 resources, request them at the same time, or obtain them from an upstream provider. It constructed named reserves for critical infrastructure, new entrants and future uncertainty. And it removed an explicit minimum allocation size. Together these measures changed size, evidence, classification and the internal partition of the pool.

The four changes were related but not interchangeable. A smaller maximum controlled the largest unit available at one decision. The utilisation and IPv6 conditions controlled what an applicant had to show. The reserves controlled which claims could reach particular slices. Removing an explicit minimum gave the rule more freedom at the small end. A published allocation policy could state all four and still fail operationally if the registry did not reveal the pool state, measurement rules and decision record that connected them.

Scarcity made the rule important, not sovereign

The practical case for rationing is not difficult to understand. A finite pool creates conflicts between present scale and future entry. Giving a larger block now may reduce repeat administration, but it removes options from later applicants. Giving smaller blocks can preserve those options, but it increases the number of interactions and may impose costs on networks that continue to grow. Reserving space protects selected future claims, but necessarily removes that space from general demand. D1 put those trade-offs into a proposed allocation sequence.

What scarcity could not do was enlarge the character of the institution administering the sequence. AFRINIC was a private technical registry and coordinator. Its useful function was to keep number-resource records unique, accurate and operationally usable, and to apply a known request process to the inventory it administered. That function can support a contractually bounded rationing rule. It does not make the registry the owner of the addresses, a government for a service region, a regulator of networks, or a tribunal for conduct.

This distinction is not an abstract objection to coordination. It is what makes coordination dependable. When the registry's role is thin and explicit, an operator knows what evidence belongs in a resource request and what consequence belongs in the ledger. When administrative terms are inflated into public authority, the consequence becomes harder to bound. A disagreement about utilisation evidence can start to sound like disobedience; a condition attached to a private service can start to sound like territorial law; a mailing-list preference can start to sound like a democratic mandate.

None of those transformations follows from the need to ration a scarce pool.

D1 should therefore be read in two registers at once. In the first, it was a serious attempt to decide who could reach which portion of a remaining inventory, through what evidential gate and in what maximum unit. In the second, every such decision remained an act of private administration. The quality of the proposal depended on whether its rule could be reproduced and corrected without pretending that the ledger created ownership or sovereignty. That is a demanding standard, but it is also a practical one.