Summary

  • AFRINIC’s account of its second meeting says the Board amended the Large IPv4 LIR annual membership fee from USD 7,000 to USD 13,000 and that the AGM unanimously approved the figures on 27 April 2005. The rise was USD 6,000, or 85.7143%, while the printed Large band remained > /16 to /14 and the initial setup fee remained USD 5,000.
  • The recovered public sequence does not show an itemized USD 13,000 proposal before the meeting, but it cannot establish that private or unarchived notice did not exist. Nor does it disclose the affected Large roster, the operative billing date, discount use, actual Large revenue, or a Large-specific cost calculation.
  • Temporary external support and AFRINIC’s stated transition toward membership-funded operations make startup self-financing a credible institutional explanation. They do not prove that the Large increase replaced grants, repaid startup assistance, reflected Large-specific service costs, or was affordable or harmful to any operator.
  • The fee financed a private technical ledger and coordination relationship. It created no taxation, ownership, confiscation, adjudication, or sovereign punishment power over number resources, networks, customers, or states.

A number that nearly doubled

The most important fact is also the least ambiguous. AFRINIC’s official account of AFRINIC-2 records a sequence in which the Board amended four figures in a proposed membership-fee schedule and the AGM then unanimously approved “these figures.” For the Large category, the recorded annual amount moved from USD 7,000 to USD 13,000. Subtraction gives a USD 6,000 increase. Dividing that increase by the old fee gives 85.7143%. The new annual figure was about 1.857143 times the old one.

That was the steepest increase among the three LIR categories whose fees rose in the reported package. Medium rose by USD 1,500, or 30%, while Extra Large rose by USD 5,000, or one third. Those two figures matter here only as boundary comparators. They establish why the Large amendment deserves its own audit; they do not turn this inquiry into an account of Medium incidence or Extra Large concentration. The subject is the Large IPv4 LIR annual price and the records needed to explain its notice, billing, revenue, and cost logic.

Two category distinctions prevent the headline from swallowing facts it does not contain. First, the archived schedules place the affected Large row in the IPv4 LIR table and print its band as > /16 to /14. The compared pre-decision and result schedules use the same band. The price changed; the published threshold did not. Second, those schedules contain separate IPv6 and End-User rows also called Large. The 2005 Large amendment examined here cannot be assigned to those separate rows merely because they share a label.

The transaction distinction is equally important. The Large initial setup fee was USD 5,000 in the captured schedule before the decision and USD 5,000 in the captured result schedule. Its change was zero. The amendment was to a recurring annual membership amount, not to the one-time setup charge. That difference matters for both interpretation and incidence. A recurring charge becomes part of a member’s continuing cash requirement whenever classification and billing bring that member within the schedule. A setup charge concerns entry or establishment and may not recur for an existing member.

Combining the two would inflate the described act and obscure what AFRINIC actually changed.

The annual figures can be expressed as monthly equivalents to show scale: USD 7,000 divided by twelve is USD 583.33, and USD 13,000 divided by twelve is USD 1,083.33. The difference is USD 500 per month equivalent. This is a comparison device only. There is no evidence in the sealed record that AFRINIC billed Large members monthly.

The schedules also displayed a 10% renewal discount. Under that stated condition, the old displayed amount was USD 6,300 and the new one USD 11,700. The difference was USD 5,400, again an increase of 85.7143%. These discounted figures are not evidence that any particular Large member qualified, paid early, or received the reduction. They define a sensitivity case. Actual eligibility, timing, and use remain unknown.

What happened in Maputo—and what the minutes do not reconstruct

AFRINIC-2 took place in Maputo on 26 and 27 April 2005. The meeting report describes more than 100 participants drawn from mixed groups: members, Board members, other Internet institutions, governments, and operators. That attendance figure must not be reused as a vote count. Event participants are not necessarily eligible AGM voters, and they are not a list of the Large LIRs that would later be billed. Three different ledgers are involved: who attended the event, who could vote in the member meeting, and who fell within the Large billing band.

The meeting report places the fee action on the second day. Its institutional sequence is specific enough to attribute responsibility at a high level: the Board amended the proposed figures, and the AGM is recorded as unanimously approving them. It is not specific enough to reconstruct the decision. The surviving account does not provide the signed motion, an individual Board vote, a voter roll, the voting denominator, the quorum, proxies, abstentions, the ballot method, or separate results for each fee line.

“Unanimously” is therefore AFRINIC’s recorded description of the AGM action, not an independently audited conclusion that every member agreed or that every affected Large member voted for the increase.

The word also carries no public mandate. An internal member approval can authorize a private corporation’s fee within the lawful scope of its corporate and contractual arrangements. It cannot make the participating membership representative of all African networks, all Internet users, or African states. It cannot create public jurisdiction by acclamation. The evidence supports an attributed corporate act; it does not support an expansion from membership procedure into continental sovereignty.

This distinction is not a semantic escape from accountability. It identifies the right accountability questions. Did the Board present an itemized proposal? Which corporate instrument governed notice? Who received it, and when? What cost or revenue objective did the USD 13,000 figure answer? When did billing switch? Those questions concern a private decision with material recurring effects. They can be asked rigorously without pretending that the fee was legislation or that AFRINIC acquired public punishment power.

The public notice trail has a gap, not a verdict

The recovered chronology begins before Maputo. On 10 March 2005, AFRINIC’s chief executive announced that registration for AFRINIC-2 was open, linked to the meeting page, and announced LIR training. The message did not state that the Large fee might become USD 13,000. An archived meeting page captured on 3 April described an open public-policy meeting and a member general meeting on Wednesday afternoon, but it did not list the fee amendment. An agenda captured on 5 April scheduled a closed AFRINIC Member Meeting from 14:00 to 17:00 on 27 April, again without itemizing the fee schedule or publishing a supporting financial paper.

Meanwhile, an official fee-page capture from 3 April displayed the Large IPv4 LIR band at the old annual figure of USD 7,000, alongside the USD 5,000 setup charge and USD 6,300 discounted amount. The meeting report dated 27 April is the first located source in the sealed record that states the new USD 13,000 figure. Put narrowly, the checked public artifacts do not close itemized advance notice of the near-doubling.

That finding must remain narrow. Public web archives are incomplete. A member email, attachment, Board paper, onsite handout, amended agenda, or other communication may have existed without appearing in the recovered set. The operative corporate notice provision from April 2005 was not located. Neither a dispatch record nor a receipt ledger is available. It follows that the evidence cannot establish that members received no notice. It also cannot establish a procedural violation, an ambush, a waiver, or valid onsite treatment of a motion.

This is a case where precise absence language matters. “The checked public pre-event artifacts did not itemize USD 13,000” is supported. “AFRINIC gave no notice” is not. The first statement describes a recovered record. The second claims knowledge of all communication channels and of the rule they were required to satisfy. The source package does not provide that knowledge.

The practical concern is still real even without a violation finding. A near-doubling is easier for an operator to budget when the amount, effective date, affected class, renewal treatment, and transition rules are stated in advance. The surviving public sequence does not supply that complete package. Yet the effect on any actual operator remains unmeasured because the record contains no operator budgets, margins, payment histories, or evidence of surprise.

The public version chain does not reveal the operative billing date

The archive becomes stranger after the meeting. An old official fee-page path still displayed the USD 7,000 Large annual fee in a capture dated 20 December 2005. The earliest located capture of a separate result-page path, dated 16 January 2006, displayed the same Large band and the same USD 5,000 setup fee, but an annual fee of USD 13,000 and a discounted amount of USD 11,700.

This establishes public version divergence. It does not establish which path AFRINIC treated as controlling, whether the old page was merely stale, when the result page was first published, or what instructions members received. A legacy URL can remain online after a newer schedule becomes operative. Conversely, a later captured page does not prove an earlier billing switch. Archive capture dates identify what was visible at those paths on those dates; they do not automatically identify effective dates.

The missing implementation documents therefore matter as much as the missing notice paper. The record has no effective-date resolution, invoice instruction, proration rule, renewal-cycle treatment, waiver schedule, or member-level invoice history. Approval on 27 April cannot simply be equated with a full-year 2005 charge. If the new price applied only at renewal, or if invoices were prorated, a model that multiplied USD 6,000 by every classified member for the entire year would overstate the year’s increment.

If the separate URLs served different audiences or one was a legacy page, their divergence might be a publishing-control problem without being a billing error.

The defensible conclusion is procedural rather than accusatory: approval, publication, and billing are separate events, and the recovered sources close only part of that chain. The act was reported in April. One old page showed the old amount in December. A separate page showed the result by January. The operative transition between those points remains unknown.

The affected category is known; the affected members are not

The schedule defines the formal affected class as IPv4 LIR membership billed in the Large band printed as > /16 to /14 when the amended annual schedule applied. That formulation contains two essential conditions: classification in the relevant IPv4 LIR band and actual application of the amended schedule. It does not include Medium or Extra Large IPv4 LIRs, the separate Large IPv6 row, the separate Large End-User row, Associate or member-only organisations, or ASN-only organisations merely by virtue of their participation in AFRINIC.

The definition is recoverable; the roster is not. The sealed record contains no dated April 2005 list of Large members, no reliable count, and no full-equivalent billed population. A 2010 registration-services presentation contains a graphical chart of LIR members by billing category across 2004 to 2010, but it lacks tabular historical values and does not provide a sufficiently precise April 2005 Large denominator. Reading a number from that graphic would manufacture accuracy the source does not support.

Even a reliable headcount would not by itself yield revenue. A member might renew after the decision, before it, or on a different cycle. Billing might be prorated. A member might qualify for the early-payment discount, receive a waiver, carry arrears, or fail to pay. Existing and newly established members may have faced different timing. Transferred organisations may have had particular setup treatment, but the record does not close how annual renewals were handled in practice. What the revenue model needs is not a guessed membership total but the number of full-equivalent amended Large billings after relevant adjustments.

Classification mechanics also remain incomplete. The printed > /16 to /14 wording tells a reader the published interval, but it does not reveal how multiple allocations were aggregated, how boundary cases were treated, whether different resource types entered the calculation, how classification was audited, or what appeal process applied. It is therefore unsafe to infer the identity of an organisation from public resource holdings and label it an affected Large payer without the historical classification ledger.

This restraint protects operators from unsupported claims as much as it protects the institution from unsupported criticism. Naming a company as affected would imply knowledge about its membership status, aggregate holdings, classification, invoice timing, and payment obligation. None of those member-specific facts is sealed. The responsible unit of analysis is the published category, not an invented list of organisations.

Revenue can be modelled only with an unresolved variable

The cleanest financial expression uses L for the number of full-equivalent annual Large billings under the amended schedule. Before discounts and timing adjustments, old schedule revenue for that cohort would be USD 7,000L, new schedule revenue USD 13,000L, and the increment USD 6,000L. If every relevant renewal received the displayed 10% discount under both schedules, the sensitivity case would be USD 6,300L before, USD 11,700L after, and an increment of USD 5,400L.

L is deliberately not called the member count. It is a billing-equivalent variable that would need to absorb effective dates, proration, renewal cycles, discounts, waivers, arrears, and collection outcomes. No numeric value for it is established. If L were zero, the schedule change would yield zero in the model. If it were positive, the modeled amount would scale linearly. Without a sealed value, any claim about actual Large-tier revenue is speculation.

The aggregate financial statement cannot fill that hole. AFRINIC’s later financial summary presented informational USD comparatives for 2005: USD 457,336 in membership fees, USD 249,904 in grants, USD 707,240 in primary income, USD 419,732 in operating expenditure, USD 10,016 in other income or expense, and a USD 297,524 surplus. Membership fees were 64.6649% of reported primary income; grants were 35.3351%. Membership-fee income was 108.9591% of operating expenditure in scale, a difference of USD 37,604, while operating expenditure was 91.7776% of membership-fee income.

Those comparisons describe institutional scale, not Large attribution. The accounts do not disaggregate membership fees by tier, disclose the number of Large invoices, identify billing periods, show discount use, or map individual collections to costs. One USD 13,000 full annual amount would equal 2.8425% of aggregate reported membership-fee income; the old USD 7,000 would equal 1.5306%; and one USD 6,000 increment would equal 1.3119%. Those are denominator illustrations only. They do not prove that a full amended billing occurred in the 2005 accounts.

The same warning applies to comparisons with broader totals. One USD 6,000 increment is 0.8484% of reported primary income, 2.4009% of grants, and 1.4295% of operating expenditure. None of these ratios allocates the increment to a named expense or demonstrates that Large caused the surplus. Multiplying by an invented number of members would convert a transparent scale exercise into a false revenue conclusion.

The meeting report contains a budget table with stated income of USD 491,000, operating expenditure of USD 360,908, and capital expenditure of USD 66,400. The arithmetic sum of the two expense lines is USD 427,308, leaving USD 63,692 against stated income; one USD 6,000 increment would be 1.2220% of that income figure. But the report’s highlights call it a 2005 budget while the table is labelled “Budget for 2006/2007.” Without a corrected record, it is not a clean decision-year denominator and cannot be used to derive the Large price.

Startup finance is context, not a cost allocation

AFRINIC did not begin in a vacuum. In March 2004, the Number Resource Organization and the emerging registry signed a startup-financing memorandum that made available a USD 100,000 envelope in five conditional payments of USD 20,000. The permitted classes included operational and capital expenses, staff and contractor costs, and directly related travel. Later payments depended on financial reporting about prior funds. The public instrument establishes the conditional design and allowable uses. It does not prove that all five stages were disbursed or received.

At AFRINIC-1 in May 2004, the official report recorded a five-category LIR fee model linked to total IPv4 allocations. It also recorded member calls for bigger LIRs to pay more and smaller LIRs less. AFRINIC’s project manager described the need to sustain operations after a two-year incubation period, with annual Board review and member endorsement of future pricing.

In its March 2005 recognition application, AFRINIC represented that membership fees were the funding base, that five categories were linked to the quantity and type of resources held, that supporting organisations would cover part or all of first-two-year costs across four locations, and that AFRINIC expected later to manage its finances from its own membership-supported revenue.

Together, those records make a transition-to-self-financing explanation credible. A startup registry receiving temporary and conditional support would reasonably need durable income for staff, capital, contractors, travel, and operations. A differentiated schedule that reduced the Associate fee while increasing larger-resource LIR fees could be understood as an effort to move more of the recurring burden toward larger categories. The reported unanimous AGM approval strengthens the case that the package had internal institutional support.

But this remains a steelman of context, not proof of the Board’s 2005 Large-specific calculation. The AFRINIC-1 discussion occurred in 2004. It cannot be treated as the later Board’s recorded motive. The recognition application stated AFRINIC’s plan; it was not an independent audit. The NRO instrument described a conditional funding envelope; it did not allocate a cost to Large LIRs or prove complete disbursement. No Board worksheet shows how USD 13,000 was derived. No per-tier budget identifies a Large revenue target.

No ledger maps the USD 6,000 increment to startup expense classes, the withdrawal of a supporter, or the replacement of grants.

A scale comparison illustrates why attribution is tempting and why it must be resisted. One USD 6,000 annual increment is 6% of the USD 100,000 conditional NRO envelope, and 16.6667 such full increments would equal the envelope. Yet numerical equivalence is not economic substitution. The funding could have arrived at different times, covered different costs, or remained partly undisbursed. Large fees could have supported ongoing operations rather than repaid startup assistance. Without receipts, phaseout schedules, per-tier forecasts, and source-to-cost mapping, the record cannot choose among those possibilities.

The unchanged USD 5,000 setup charge adds an important audit question. The institution left the price explicitly labelled as initial setup unchanged while sharply increasing recurring annual membership. That does not prove recurring costs had grown, nor does it prove the opposite. It isolates the decision as a recurring-price adjustment and invites a request for the recurring-cost or burden-allocation model behind it. Was the intended basis service workload, resources held, ability to pay, a cross-subsidy, the end of external support, a negotiated package, or some combination? The sealed record cannot answer.

What operators bore—and what cannot be said about harm

For a Large LIR that received a full-equivalent amended annual billing without discount, the nominal schedule burden was USD 6,000 higher. Under the fully discounted sensitivity, it was USD 5,400 higher. Those figures describe the direct cash mechanism. They do not establish affordability, regressivity, discrimination, margin compression, customer pass-through, reduced investment, or service cuts.

Those downstream outcomes require operator evidence: revenue, costs, margins, customer counts, investment plans, invoices, payment dates, and managerial responses. None is available in the sealed set. A USD 6,000 increase may be material to one network and modest to another, but that range cannot be resolved by assuming that all organisations within a holdings band have the same financial capacity. Nor does a holdings-linked category automatically measure the registry workload that a member creates.

The member’s underlying economic value also needs conceptual separation from the ledger charge. Networks, operators, and customers create and use connectivity. AFRINIC coordinates unique number-resource records and related administrative services. Its proximity to the ledger does not make it owner of routed resources or of the businesses built with them. The fee finances a private registry relationship; it is not a purchase price for sovereign title and does not give AFRINIC a claim to the value created downstream.

The same boundary governs enforcement language. A lawful private corporation may have contractual or corporate remedies, and their exact scope would need separate proof. But this fee resolution created no inherent power to tax, prosecute, adjudicate, seize, confiscate, or punish as a sovereign. ICANN’s recognition of AFRINIC as a Regional Internet Registry on 8 April 2005 did not approve the USD 13,000 amount and did not confer such powers. Recognition describes a coordination role; it is not a transfer of governmental jurisdiction.

The strongest public conclusion is an audit frame

The Large amendment is not obscure in its arithmetic. It is USD 7,000 to USD 13,000, an additional USD 6,000 and 85.7143%. The published band remained > /16 to /14; the setup fee remained USD 5,000; the separate IPv6 and End-User Large rows were not the object of this act. The Board-to-AGM sequence and reported unanimity are clear as AFRINIC’s own account.

What is missing forms a coherent ledger rather than a cloud of generalized suspicion. Notice needs the operative rule, an itemized dispatch, recipients, attachments, and receipt or waiver evidence. Affected-member analysis needs a dated roster, operational classification rules, and full-equivalent amended billings. Implementation needs the effective date, renewal transition, proration, controlling publication path, and invoices. Revenue needs discounts, collections, arrears, and tier disaggregation. Cost allocation needs the Board model, startup-support receipts, phaseout terms, expense mapping, and per-tier forecasts.

Impact needs operator evidence.

Until those records appear, two symmetrical overclaims should be rejected. The increase cannot be declared unfair, regressive, harmful, procedurally invalid, or unjustified from the surviving evidence. It also cannot be declared cost-based, fully noticed, accurately implemented, or harmless. The credible startup self-financing account belongs in the analysis, but so does the inability to tie USD 13,000 to the Large cohort’s costs or to reconstruct what members saw and paid.

The enduring principle is modest. A ledger coordinator may charge for coordination through a private relationship. The sharper and less explained the recurring price movement, the stronger the case for a clean supporting ledger. That ledger would not confer sovereignty; it would make a corporate price decision intelligible.