Summary

  • On 6 March 2009, ICANN Board resolution 2009.03.06.02 ratified the global policy for allocating IANA’s remaining IPv4 address space and instructed staff to implement it. AFRINIC’s archive records AFPUB-2009-v4-001 as implemented on that same date. These are connected but distinct acts: ICANN made the global ratification decision, while AFRINIC recorded the status of a proposal that its own region had adopted on 13 August 2008.
  • The policy reserved one /8 for each of the five recognized RIRs. Ordinary IANA allocation continued until an eligible RIR request either could not be fulfilled from the non-reserved pool or could be fulfilled only by emptying it. At that trigger, the requesting RIR would receive the remaining ordinary units, M, and every RIR would receive one reserved /8. The formula therefore combined a symmetric floor with a request-dependent remainder.
  • Equal quantities at the RIR tier did not establish equal operator need. The rule did not allocate directly to an ISP, enterprise, school, access network, country, population or end user. It did not measure infrastructure, historical holdings, documentation capacity, timing or deployment constraints. It reduced terminal timing risk among five registries; it did not settle distributive justice downstream.
  • AFRINIC’s role remained that of a private technical coordinator and bookkeeper. Recording allocations, preserving uniqueness and participating in a global inventory protocol did not give it ownership of address space, sovereignty over a service region, or legislative, regulatory, police, prosecutorial, judicial, punitive or confiscatory power.

Analysis

One date, several institutional acts

The importance of 6 March 2009 lies partly in what happened and partly in keeping straight what did not happen. The ICANN Board adopted resolution 2009.03.06.02, ratifying the proposal as a Global Policy and instructing ICANN staff to implement it in time for its provisions to take effect. AFRINIC’s policy archive assigns AFPUB-2009-v4-001 an implemented status on the same date. That alignment marks the completion of the agreed global-policy chain and the activation of an inventory rule at the IANA-to-RIR layer.

It was not AFRINIC’s first adoption of the proposal. The African regional process had begun under the older reference afpol-v4gp200802. AFRINIC’s record says consensus on an updated proposal was reached at AFRINIC 8 on 5 June 2008, followed by a fifteen-day last call from 23 June to 8 July. ICANN’s policy tracker records adoption by the AFRINIC Board on 13 August 2008. The archive’s top-line “Submitted” date of 6 March 2009 is therefore ambiguous metadata, not a sound basis for moving the proposal’s introduction or regional adoption forward by a year.

Nor was 6 March the date on which IANA distributed the final five /8s to the registries. The implementation created and governed a reservation and a conditional terminal procedure. A status report ten days later said that, out of 32 unallocated unicast /8s, 27 remained in the ordinary free pool and five had been reserved, one for each RIR. Reservation changed the inventory available to ordinary requests. It did not itself constitute the later terminal allocation, still less an allocation to operators.

That distinction matters because a compressed account can turn three different events into one misleading sentence. Regional adoption in 2008, global ratification and recorded implementation in March 2009, and conditional terminal allocation were separate points in the chain. Collapsing them would exaggerate AFRINIC’s authority, misstate ICANN’s act and erase the trigger that made the final distribution contingent rather than immediate.

The narrow event can therefore be stated precisely: 6 March 2009 was the global ratification and recorded implementation date, not the date AFRINIC first adopted the proposal and not the date IANA distributed the final five /8s.

How a regional proposal became a global inventory rule

The global process was not a single vote by a global electorate. It was a sequence of institutional steps created for policies that must be identical across all RIRs and require action by IANA or ICANN. Under that arrangement, regional approval supplies the common input; the NRO and ASO perform coordinating and process-review functions; and the ICANN Board supplies ratification within the agreed mechanism.

The text itself emerged from a period in which two broad approaches shared the idea of reserving N /8s per registry but differed over the value of N. By February 2008, a unified proposal fixed N at one. That bounded fact matters because it identifies the formula ultimately implemented. It does not require reopening the earlier AFRINIC-7 response record or treating that earlier episode as evidence about the legitimacy or distributive effect of the final policy.

AFRINIC introduced the unified proposal into its process on 16 February 2008. Its recorded consensus, last call and Board adoption followed. The other regional processes proceeded on their own tracks, and APNIC’s Executive Council completed the last of the five RIR adoptions on 20 November 2008. Identical adoption mattered because the policy was supposed to control one shared IANA inventory. Five inconsistent formulas would not have provided a single terminal rule.

On 3 December 2008, the NRO Executive Council submitted the coordinated proposal to the ASO Address Council. The ASO Address Council adopted it after process review on 8 January 2009. Its chair forwarded the proposed global policy to the ICANN Board on 4 February, while ICANN circulated it to its advisory committees and supporting organizations. ICANN then opened a final public-comment period from 5 to 26 February.

The Board resolution records one comment by the deadline, and that comment supported the proposal. One is a definite number in the record, but it has no representative denominator attached to it. It cannot responsibly be converted into proof that operators, users, countries, populations or the world at large consented. It shows that the recorded final call drew one supporting comment; no more representative conclusion follows.

On 5 March, the ASO Address Council sent advice in full support to the ICANN Board. On 6 March, the Board found the policy predictable, ratified it and instructed staff to implement it. AFRINIC’s same-day implementation label belongs at the regional archive end of that connected sequence. AFRINIC did not issue the ICANN resolution; ICANN did not retroactively become the author of AFRINIC’s 2008 regional adoption.

The authors named on AFPUB-2009-v4-001—Roque Gagliano, Francisco Obispo, Haitham EL Nakhal, Didier Allain Kla, and the named JPNIC IPv4 countdown policy team—contributed the policy text. Authorship identifies contribution. It does not, by itself, establish a mandate from every operator or prove that the policy’s consequences were fair. The same discipline applies to every actor in the chain: the evidence of a procedural role is not evidence of a larger political authority.

The five RIR communities and boards processed the same proposal; the NRO Executive Council submitted it; the ASO Address Council reviewed and forwarded it; the ICANN Board ratified it; and ICANN staff acting through the IANA function were to carry it out. This allocation of tasks is the control surface. It prevents the convenient but false claim that any one participant unilaterally legislated for the Internet.

The two phases and the exact trigger

AFPUB-2009-v4-001 is easiest to misunderstand when reduced to “one last /8 each.” That phrase captures the reserved floor but omits both the ordinary-allocation phase and the residual M allocation. The actual mechanism has two phases joined by a request test.

Let R be the number of recognized regional Internet registries. In March 2009, R equalled five. Let N be the number of identical terminal /8 units reserved for each registry. The unified policy fixed N at one. The reservation step was therefore R multiplied by N: five registries times one /8, producing five reserved /8s. Those units were removed from the inventory available for ordinary IANA allocation.

The existing-policy phase did not stop on 6 March or on a predetermined calendar date. IANA continued ordinary allocation from the non-reserved pool. Let Q be the size of an RIR’s eligible request and A be the available non-reserved IANA allocation units when that request was evaluated. The transition occurred under either of two conditions: Q could not be fulfilled from A, or Q could be fulfilled but doing so would leave A at zero.

The second branch is important. A request that exactly emptied the ordinary pool activated the exhaustion phase even though it could technically be satisfied. A description that says only “the trigger occurred when a request could not be filled” leaves out half of the operative test. A description that says “the rule began when the calendar reached exhaustion” removes the request evaluation altogether. Both simplifications obscure how the system was designed to behave.

The request tested at that boundary was the last request IANA accepted under the existing policy. Once the terminal condition was reached, IANA was to notify the NRO. It would allocate M—the remaining available ordinary units after evaluation of the last request—to the RIR that had submitted that request. It would also allocate one reserved /8 to each of the five RIRs.

M is not a decorative algebraic detail. It reveals an asymmetry inside the symmetric rule. Each RIR had the same one-/8 reserved floor, but one RIR, the final requester, could also receive the remaining ordinary units. Total receipts at the terminal operation therefore did not have to be identical. The rule guaranteed five equal reserved quantities and separately disposed of a request-dependent remainder.

The value of M is not established by the March 2009 implementation record. It depended on the pool and request at the later trigger. Assigning it a value would convert a formula into an invented outcome. The responsible account preserves M as unknown and keeps the terminal sequence conditional.

The policy also provided that IANA should reserve one /8 for a newly recognized RIR at the time of recognition. That clause describes how the formula would respond to a new registry. It does not establish that a sixth RIR appeared in 2009, and no such event should be inferred from the rule.

The distinction between reservation, allocation and use runs through the entire mechanism. Reservation ring-fenced inventory. Allocation at the IANA-to-RIR layer would change which registry received which block under the global formula. Downstream RIR allocation would address applications from members or operators under regional rules. Operational use would depend on an issued prefix being deployed and routed. These are not interchangeable stages.

The arithmetic—and what the arithmetic cannot show

A /8 fixes eight bits of a 32-bit IPv4 address, leaving 24 variable bits. Its numerical size is therefore 2 to the power of 24, or 16,777,216 addresses. Five /8s represent 83,886,080 numerical addresses. That is a measure of block size, not a promise that every numerical value is independently usable by an operator or end user.

The 16 March 2009 status report provides a clear inventory snapshot. It stated that there were 32 unallocated unicast /8s: 27 in the ordinary free pool and five reserved for the RIRs. The reserved portion was five thirty-seconds of that count, or 15.625 percent. This ratio describes the split of the unallocated-unicast /8 inventory at that moment. It says nothing about later operator distribution, regional utilization or who would ultimately benefit.

The numbers are material because reservation changed the risk faced by the institutions at the tail of the pool. Without a protected floor, a large request arriving late in the ordinary process could consume the remaining inventory and leave another RIR with nothing from the terminal stock. Ring-fencing one /8 each reduced that arrival-order risk and made a minimum quantity predictable for every recognized registry.

Predictability has operational value. A regional administrator that knows one /8 is protected can plan for a terminal phase without treating every ordinary request by another RIR as a threat to its entire remaining opportunity. The rule thus dampened one kind of global race. It converted an uncertain contest over the last available units into a common floor plus a specified treatment of the final ordinary request.

Yet the arithmetic cannot identify a deserving operator. It cannot determine whether a new ISP, an established carrier, a cloud provider, a school, a small enterprise or an access network has the stronger claim. It cannot measure whether an applicant can document need, finance deployment, acquire equipment, reach users or satisfy a regional rule. A block count is an inventory fact, not a welfare function.

Nor does 83,886,080 numerical addresses translate into 83,886,080 equivalent opportunities. The numerical total does not reveal downstream policy, infrastructure, administrative capacity, historical stocks or the timing of demand. It would be a category error to turn a precise binary calculation into a claim about equal economic benefit.

Equality of what, and among whom?

The formula’s unit of equality was the RIR. Five recognized registries each received a protected claim to one /8 when the terminal procedure eventually ran. It was not a formula assigning equal quantities to countries, equal chances to networks, equal service to users or equal capacity to communities. Those actors were downstream from the instrument.

This is more than a semantic distinction. An equality claim is only as sound as its unit and denominator. “One each” sounds distributive until the noun is supplied. One /8 for each registry is institutional symmetry among five coordinating bodies. It does not become equality among African, Asian, European, Latin American, Caribbean or North American operators merely because the registries have regional names.

The policy’s own rationale recognized that terminal needs would vary by region. That recognition helps explain why the global text did not attempt an operator-level needs schedule. Regional processes were left to decide how their allocated inventory would be distributed. The formula solved a common-pool problem at one layer and deliberately declined to solve heterogenous access problems at the next.

That design choice can be sensible without being distributively complete. AFRINIC, APNIC, ARIN, LACNIC and RIPE NCC could receive the same reserved prefix size even though their historical stocks, populations, network scale, demand growth, documentation capacity and downstream rules differed. The equal floor suppressed those differences rather than resolving them.

For operators, access would still depend on variables outside the global rule: eligibility, the evidence an applicant could produce, when it applied, the holdings and infrastructure already available to it, and the regional policy governing the request. A provider with capital and administrative capacity could be situated differently from a school or small access network even if both operated under the same RIR. Nothing in one equal registry floor measures or eliminates that distance.

Institutional equality can coexist with unequal operator access. That proposition does not prove that the policy harmed a particular operator, favoured a particular country or produced any measured later outcome. The available record does not supply those results. It simply establishes that the formula contained no operator-level variable capable of proving equality among those actors.

The error runs in both directions. It would be wrong to celebrate the formula as proof of development justice, but it would also be wrong to condemn the equal floor merely because it was coarse. A global inventory mechanism may reasonably use a simple institutional unit precisely because it lacks the legitimacy, data and competence to rank every social need. The correct judgment depends on keeping its claims proportional to its function.

The strongest case for a simple floor

The best defence of the policy begins with the conditions at the common layer. A terminal pool needed a rule that all five RIR systems could understand and that IANA could execute without becoming a global tribunal of social and economic demand. One /8 per recognized RIR was simple, auditable and resistant to endless recalculation.

The floor protected each regional administrator against the possibility that the final large on-demand request would leave it with no terminal block. It also preserved room for different regional policies. If needs varied, as the policy acknowledged, then moving downstream decisions closer to the regional processes avoided forcing IANA to choose among incomparable claims from thousands of networks.

A weighted global formula would have required a moral and statistical theory. Population might favour one result, subscriber count another, historical holdings a third and projected growth a fourth. Each denominator could become stale or strategically reported. Disputes would require audit, interpretation and appeal. The more the global layer tried to produce social equality, the more discretion it would accumulate over regional outcomes.

On that account, equal /8s were not a declaration that regions were equal. They were a refusal to turn IANA into a welfare allocator. The simplicity of the formula constrained the common institution’s power, while the reservation prevented a pure race in which request timing could determine whether an entire RIR received any terminal floor at all.

This defence gets several things right. Predictability and minimum service continuity for each registry were legitimate operational objectives. An identical prefix was easier to audit than a synthetic need score. Need did vary within as well as across regions. And separating the common inventory rule from regional distribution kept the global function thinner.

But the defence supports only a limited conclusion. It justifies a registry floor, not a claim of equal operator need. Indeed, its premise is that the common layer could not responsibly measure those needs. The policy is most defensible when advocates resist decorating it with larger claims about regional entitlement, political representation or fairness to end users.

The same is true of the M remainder. A simple slogan says that the last five /8s were split equally. The formula actually paired five equal reserved units with a residual allocation to the final requesting RIR. This did not invalidate the floor, but it did mean that the total terminal operation was not simply five identical receipts. Transparency about M strengthens the rule’s auditable character; hiding it weakens the very simplicity argument on which the policy relies.

A private coordination chain, not a grant of sovereignty

The global-policy process was an agreement among private coordinating institutions about a shared technical inventory. The ASO Memorandum of Understanding defined the kind of policy that required agreement among the RIRs and ICANN and assigned ratification to the ICANN Board. That framework gave the participants a process for coordinated action. It was not a treaty, a statute or a delegation of sovereign power.

AFRINIC’s exact role should be described without either minimizing or inflating it. It hosted the African regional process, recorded consensus and last call, adopted the identical proposal through its Board in 2008, and records AFPUB-2009-v4-001 as implemented on 6 March 2009. Those were consequential coordination acts. They did not make AFRINIC Africa itself.

AFRINIC remained a private technical coordinator and bookkeeper. It had no sovereign, legislative, regulatory, police, prosecutorial, judicial, punitive, confiscatory or ownership power. Its service region was an administrative geography, not a sovereign constituency. A registry’s record can coordinate the uniqueness and status of number resources without manufacturing political title to them.

The same boundary applies to ICANN. Its Board resolution was a corporate decision within the agreed global-policy mechanism. It directed staff action over the IANA inventory. It was not a public-law referendum, a property deed, a criminal code or proof that every affected operator had delegated governmental authority to the Board.

The NRO Executive Council coordinated submission. The ASO Address Council reviewed the regional-process record, adopted and forwarded the proposal, and advised the Board. These were procedural functions under the established chain, not a legislature, court or source of coercive jurisdiction. IANA reserved inventory and was to apply the trigger formula; it did not adjudicate operator welfare or acquire ownership of routed networks.

The thin-function distinction is essential. A registry may record allocations, coordinate uniqueness, preserve accuracy and support continuity. Those actions can be technically indispensable. Their indispensability does not entitle the registry to rule the people who use networks, punish holders, confiscate resources or convert the map of service responsibilities into political property.

This boundary also prevents an opposite misunderstanding: rejecting sovereignty does not make the registry’s acts meaningless. A bookkeeper can make decisions that materially alter the availability and timing of an inventory. A ledger can be authoritative for coordination without being a constitution. The right response is not to deny the institution’s technical power but to describe that power precisely and confine the conclusions drawn from it.

Global allocation coordination likewise creates no regional ownership. The fact that one /8 was reserved for each RIR did not vest a continent with title, establish a popular mandate or authorize the registry to treat already-issued resources as political assets. It arranged the next step in a shared administrative chain.

Where advocacy organizations participate in debate, their roles must remain equally bounded. NRS may advocate, research, convene and represent members who have expressly authorized it. It does not operate a registry, RPKI, WHOIS or RDAP; administer appeals or settlements; run elections; hold custody; or provide registry continuity. Nothing about the final-/8 policy transfers those functions to it or makes advocacy a source of sovereign authority.

What the rule changed

The immediate change was to the IANA inventory. Five /8s were removed from the ordinary pool and protected for one-per-RIR terminal allocation. Ordinary requests continued against the remaining non-reserved inventory until one met either branch of the trigger. At that point, the policy specified notification, the M remainder and the five reserved allocations.

This redistributed terminal timing risk among registries. Under pure on-demand allocation, the size and arrival of the last requests could determine whether some RIR received none of the remaining globally unallocated inventory. Under AFPUB-2009-v4-001, every recognized RIR had a floor insulated from that race. The final requester still occupied a distinct position because it could receive M.

The rule did not alter historical holdings. It did not decide how an RIR should allocate its terminal stock to an operator. It did not finance equipment, train staff, lower documentation burdens or create routes. It did not measure end-user need. It did not allocate national shares or grant a government title over addresses associated with networks in its territory.

The reservation nonetheless raised the importance of regional policy. Once a registry knew it had a protected terminal /8, the questions of eligibility, evidence and distribution moved decisively to the downstream layer. This was not an accidental loophole; the policy’s rationale acknowledged different regional needs and left those choices to the regions. A thin global rule can therefore make the quality of a lower-layer rule more consequential without itself answering the lower-layer questions.

The stakes are best understood as option value rather than as a proven later outcome. A guaranteed floor gave each RIR a basis for planning. It reduced uncertainty about receiving zero from the terminal stock. But the option could have different significance for different operators depending on downstream access. The record supports the mechanism, not a claim about which operator ultimately gained or lost.

The formula also shaped public understanding. “One final /8 for each region” can sound like a continental settlement, even though the actual recipient class was five registries. If repeated without the administrative qualifier, the phrase can turn service geography into an imagined property boundary. If described as an IANA-to-RIR inventory rule, it remains intelligible without acquiring a political meaning it never contained.

Three questions the formula deliberately kept apart

The first question was one of uniqueness and inventory: how should the remaining ordinary pool be protected against a final request sequence that could leave a recognized registry without a terminal block? The policy answered that question directly. It reserved five units, defined the transition test and assigned the remainder. Its variables were suited to a ledger because they described registries, requests and available blocks.

The second question was one of downstream administration: which applicants should receive portions of a registry’s inventory, in what sizes, on what evidence and at what times? The global policy did not provide those answers. Leaving them to regional processes recognized variation, but it also meant that the one-/8 guarantee could not be used as evidence that any particular applicant would qualify or receive service.

The third question was one of social outcome: whose connectivity, opportunity or economic position would improve, and by how much? Nothing in R, N, Q, A or M measured that outcome. Those variables could produce an auditable terminal allocation without revealing whether downstream access was broad, narrow, fast, slow, inexpensive or burdensome. A reader who wants an answer to the third question needs evidence from beyond the formula, not a more enthusiastic interpretation of it.

Keeping the questions separate avoids two forms of institutional inflation. One is analytical: treating a successful inventory operation as proof of social equality. The other is constitutional: treating the power to maintain the inventory as authority to govern the people affected by networks. Neither follows from the importance of the coordination task.

The separation also clarifies accountability. IANA could be assessed against the reservation and trigger. The organizations in the global-policy chain could be assessed against their stated procedural roles. A regional registry could be assessed against the downstream rules it actually applied. Operators could be assessed against the evidence and operational obligations relevant to their own requests. No actor needed to be imagined as master of the entire system for each actor’s narrower conduct to remain visible.

That is why the policy should not be dismissed as merely clerical. A thin rule can materially alter timing and options while remaining thin in authority. The correct standard is exactness: recognize what the ledger changed, demand evidence for any downstream claim, and refuse to turn administrative reach into ownership.

Facts that the record does not establish

The policy record is detailed about dates, actors, variables and inventory, but it does not answer every question that later readers might bring to it. The gaps are substantive limits, not invitations to guess.

The AFRINIC archive’s “Submitted: 06 March 2009” label conflicts with the chronology showing introduction in February 2008 and Board adoption in August 2008. The safe conclusion is that the label’s current meaning is uncertain. It should not be used as a new first-submission date or as evidence of a second regional adoption. An apparent inconsistency in the archive’s label for a separate 2007 event is outside this policy’s narrow act and need not be interpreted.

The record does not provide the AFRINIC Board’s 13 August 2008 tally, the individual votes, meeting minutes or a denominator. It records adoption, not the representative reach of that decision. Similarly, the ICANN Board resolution’s one supporting public comment has no disclosed population, membership, operator or end-user denominator. Neither institutional record can be inflated into evidence of universal consent.

The record does not quantify regional operator demand, country need, applicant need, affordability, end-user benefit or later distribution outcomes. It does not establish who ultimately won or lost, whether any particular network was harmed, or whether one region used its inventory better than another. Assertions on those questions would require evidence beyond the implemented formula.

M remained a variable in March 2009. The implementation record did not specify its later value. The final five /8s had been reserved, not yet conditionally allocated. No later ceremony, depletion timeline, transfer market, queue, recovery practice, lease, revocation dispute, litigation, insolvency event or governance crisis belongs in the explanation of this act.

No event-specific material for AFPUB-2009-v4-001 is available here from nrs.help, heng.lu, larus.net or BTW.Media. The Heng materials bear on doctrine—the limits of registries, ownership language and institutional equality—not on the event’s dates, votes or results. The absence of an event-specific item in this record is not proof that no view or document exists elsewhere.

The boundary around later events is especially important because hindsight can make the 2009 formula appear to have determined everything that followed. It did not. Later regional soft-landing texts, account-level limits, queues, transfers, depletion and market behaviour would involve separate rules and evidence. Importing their outcomes into the March 2009 act would confuse an upstream allocation formula with downstream histories.

The defensible conclusion

AFPUB-2009-v4-001 solved a bounded coordination problem. It ensured that each of five regional registries would have one protected terminal /8 and specified the request conditions that would end ordinary allocation. It preserved the residual M for the last requesting RIR and therefore combined institutional symmetry with a limited request-dependent asymmetry.

Its virtue was predictability. Each RIR was protected from receiving zero merely because another registry’s eligible request arrived first or exhausted the ordinary pool. IANA could execute a deterministic rule without ranking countries, populations or operators. The common layer stayed comparatively thin.

Its limit was equally clear. One /8 each was not a census of need. It did not account for historical holdings, network scale, infrastructure, documentation capacity, timing, deployment or regional eligibility. It guaranteed inventory to institutions, not equal access to the people and networks downstream.

The clean reading therefore resists two temptations. The first is to call a simple registry floor proof of distributive justice. The second is to treat coordination over a scarce inventory as ownership or sovereignty. Neither follows. AFRINIC remained a private coordinator and bookkeeper, not a sovereign or owner; its service region remained an administrative area, not a political constituency.

The policy’s defensible achievement was predictable, thin coordination at the boundary between IANA and five registries. Its formula should be read neither as a census of need nor as a constitutional settlement: one /8 per registry was an institutional floor, while downstream access and operator need remained separate regional questions.