Summary

  • AFRINIC’s meeting report records that its Board amended the Medium LIR annual fee from USD 5,000 to USD 6,500 on 27 April 2005 and that the AGM unanimously approved the amended figures. The change was USD 1,500, or 30%, but the record does not disclose a separate Medium vote, the electorate, an effective invoice date, the affected-member count or the Board’s calculation.
  • The later published schedule placed Medium in a /18-to-/16 band. That supports an inference that the fee allocated costs through a holdings category, while the flat USD 6,500 price inside a fourfold address-count range shows that it was not exactly proportional to holdings. The public record does not close either Medium-specific service burden or the financial necessity of precisely USD 6,500.

The consequential middle row

The revealing feature of AFRINIC’s 2005 fee change is not that Medium received the largest increase. It did not. The revealing feature is that a category covering the middle of the LIR schedule acquired a single, higher recurring price without a published account of what, within that category, the extra money was meant to follow.

The official AFRINIC-2 meeting report records a short sequence on the second day of the Maputo meeting, 27 April 2005. The Board made four amendments to fee figures. The Medium annual fee moved from USD 5,000 to USD 6,500. The report then says the AGM unanimously approved the amended figures. Those recorded acts close the basic event and the arithmetic: USD 6,500 minus USD 5,000 is USD 1,500; USD 1,500 divided by USD 5,000 is 30%; the new standard fee was 1.3 times the old one.

What the report does not supply is just as important. It does not reproduce a Medium-only motion or show whether the four changes were bundled. It does not publish the voter roll, quorum, denominator, proxies, abstentions, ballot method or category representation. It does not give the Board’s reasons, alternatives or model. It does not identify the first invoice, an effective date, a transition rule or the number of LIRs that actually fell within Medium. “Unanimously approved” therefore describes the outcome AFRINIC recorded.

It cannot be enlarged into proof that every member voted, that every Medium LIR supported the rise, or that all operators and users in the service region consented.

This distinction is not a technical reservation appended to an otherwise complete story. It determines which story the evidence can sustain. Three possible allocation rationales are available. The price might have followed registry footprint, using resource holdings as a proxy for the share of common cost a member should bear. It might have followed category-specific service burden, with Medium members causing a measurable quantity of requests, support or database work. Or it might have been chosen to fill a quantified financial need during AFRINIC’s transition from incubation support to self-finance.

The surviving record strongly supports the first as a design inference, establishes only a common service surface for the second, and supplies general institutional context but no exact derivation for the third.

What the publication trail fixes in place

The meeting report is not the only relevant instrument, and each document performs a different evidentiary job. A March 2004 document labelled Draft01a printed a Medium standard annual fee of USD 5,000, a discounted figure of USD 4,500 and a USD 3,000 setup fee. Because it is expressly a draft and its printed Small and Medium ranges overlap, it cannot by itself establish the final pre-vote thresholds, who was classified where, or whether anyone moved between categories. The AFRINIC-2 report independently confirms the USD 5,000 starting figure, so the basic price change does not depend on treating the draft as a final schedule.

A later official fee page, version 20050400 and captured on 16 January 2006, shows the resulting Medium row for the 2005/2006 schedule. It publishes a /18-to-/16 range, a USD 3,000 initial setup fee, a USD 6,500 standard annual membership fee and USD 5,850 under the stated 10% early-renewal condition. This page is strong evidence of what AFRINIC subsequently published. Its capture date does not prove that the page was unchanged from 27 April, that the new rate was invoiced immediately, that a particular LIR qualified for the discount, or that any invoice was paid.

The distinction between adoption, publication and collection prevents false certainty. Approval on 27 April is not automatically a full-year 2005 billing assumption. A row on a later archived page is not an invoice ledger. A percentage discount is not evidence of renewal behaviour. Without the effective-date resolution, renewal cycles, proration rules, waiver records, arrears and cash receipts, the standard and discounted figures show possible contractual charges, not realised Medium revenue.

The archive itself also has limits. The selected present-day URLs for the AFRINIC-1 report, the AFRINIC-2 report and the 2006 financial statements returned 404 responses by the research cutoff. Complete archived captures were recovered; the AFRINIC-2 report was text-extracted and page 8 visually checked. That recovery preserves the examined contents. It does not make a dead current URL live, restore an unpublished annex or fill a missing denominator.

Rationale one: holdings and registry footprint

The strongest explanation in the record is that Medium was a holdings-banded charge. AFRINIC’s founding meeting report described five LIR categories as determined by total IPv4 allocated. Its March 2005 recognition application described five categories linked to the quantity and type of resources held. The resulting fee page then placed Medium in the printed /18-to-/16 row. Taken together, these are affirmative evidence of a resource-holdings proxy, not merely an inference from the word “Medium.”

But a band is not a meter. In a simple single-prefix illustration, a /18 contains 16,384 IPv4 addresses, a /17 contains 32,768 and a /16 contains 65,536. The endpoints differ fourfold. Applying the same USD 6,500 standard annual price across that row produces nominal fee-per-address ratios of about USD 0.396729 at /18, USD 0.198364 at /17 and USD 0.099182 at /16. The ratio at the /18 endpoint is four times the ratio at the /16 endpoint.

This calculation establishes the shape of the published row, not the economics of any member. Addresses are not metered service units. The calculation does not show routed use, customers, turnover, profitability or ability to pay. Nor does the table disclose how AFRINIC aggregated multiple prefixes, treated resource types, rounded a total or resolved overlap. A member-level classification ledger for April 2005 was not located. The single-prefix exercise therefore illustrates within-band incidence; it does not identify an affected operator or reconstruct actual classification.

The fee was holdings-linked in a coarse sense and non-proportional in a finer sense. Crossing into Medium placed a member under one price, while movement inside Medium from the lower to the upper endpoint did not raise that standard price. Such a staircase can be administratively predictable: a published band tells a member the charge associated with its class. It can also create threshold pressure. A small increment in holdings near a category boundary may be associated with a discontinuous change in annual cost even though the network’s underlying expansion is incremental.

That threshold mechanism is real as a property of a banded design, but its historical effects remain unobserved here. The checked record does not name a member that crossed into Medium, show a request delayed because of a fee boundary, document a reclassification dispute, or publish proration treatment. The predecessor draft’s overlapping printed rows cannot be used to claim that the 2005 schedule reclassified anyone. To move from structural incentive to observed behaviour would require the before-and-after member ledger and contemporaneous request records.

The middle row also matters because averaging can conceal its internal distribution. Two LIRs could face the same USD 1,500 rise while sitting at quite different points within the printed band. The schedule’s classification logic makes both “Medium”; the flat fee gives no further recognition to their location inside it. If holdings are intended as a proxy for capacity or benefit, the proxy is discontinuous at category edges and deliberately blunt within a category. The record supports that distributional mechanism. It does not tell us whether AFRINIC considered the bluntness desirable, tolerable or merely easy to administer.

Rationale two: a Medium-specific service burden

AFRINIC did provide services for which a private organisation could charge. Its March 2005 LIR setup guide described a service agreement, payment before service, resource requests, registry identifiers, member assistance and training. The training surface included Whois, policy, reverse delegation, the routing registry and AS numbers. The fee page characterised the annual amount as a charge for service and membership. These materials defeat any suggestion that the invoice paid for literally nothing.

They do not, however, connect Medium to a distinct amount of work. The checked documents do not disclose Medium ticket volume, hostmaster hours, allocation requests, database changes, reverse-DNS work, training attendance, support contacts or marginal cost. They do not compare a /18 holder with a /16 holder, or Medium with the categories on either side, on any operational cost driver. There is no service ledger showing that a member’s category predicted how frequently it called on staff or how costly its records were to maintain.

It follows that service access and service causation must be separated. Paying the annual fee could be a condition of access to the common LIR service surface. That fact does not prove the extra USD 1,500 measured an extra quantity of service. Still less does a 30% price increase establish a 30% increase in Medium workload. No such workload increase is closed by the record.

A service-burden rationale remains possible. Medium members might have generated more hostmaster work, more complex requests, greater training demand or more costly support than smaller members. If a contemporaneous cost-driver study showed that pattern, USD 6,500 could be partly service-cost reflective. The decisive point is that the antecedent is missing. The public record leaves Medium-specific service burden unclosed and leaves the causal relationship between the category and the extra USD 1,500 unclosed.

This gap is analytically different from disagreement over fairness. One need not decide whether a holdings proxy is fair to notice that it is not proof of service consumption. Holdings can be chosen as a way to distribute common fixed costs, as an assumed indicator of ability to pay, as a measure of institutional benefit, or simply as an established billing convention. Those are different principles. Without the Board paper or a cost allocation, the fee table does not tell us which principle carried the amount from USD 5,000 to USD 6,500.

Rationale three: financial necessity

The case for general financial need is better documented than the case for Medium-specific service burden. AFRINIC-1 recorded a sustainability discussion during the registry’s formation. Management referred to self-sustainability after a two-year incubation period and to annual fee revision with member endorsement. Entities called for bigger LIRs to pay more and smaller ones less. AFRINIC’s recognition application said membership fees were the funding base, that supporting organisations covered part or all of the first two years’ costs, and that the registry was expected to become self-financing after incubation.

There was also an announced NRO startup-support envelope of USD 100,000 in five conditional USD 20,000 instalments. The available record establishes the announced support and its conditions, not that every instalment was disbursed or that the Medium increase replaced a particular payment. ICANN’s recognition of AFRINIC on 8 April 2005 supplies institutional chronology. Recognition did not set the fee, validate its amount or confer public authority over the operators whose records AFRINIC coordinated.

Later financial statements provide useful scale but not the missing derivation. In informational USD comparatives for 2005, AFRINIC reported USD 457,336 in membership-fee income, USD 249,904 in grants and USD 707,240 in primary income. It reported USD 419,732 in operating expenditure, USD 10,016 in other income or expense and a USD 297,524 surplus. Membership fees were about 64.6649% of primary income; grants were about 35.3351%. Reported fee income exceeded operating expenditure by USD 37,604 and equalled about 108.9591% of it.

Those numbers show that membership fees were material to the startup registry. They do not show how much Medium supplied. There is no Medium member count, billed-revenue total, collections line, discount total, waiver record, arrears figure or bad-debt amount. The USD conversions were provided for information and had no official status relative to the audited Mauritian-rupee statements. The accounts are also ex post and aggregate, while the question is why the Board selected a category price at an earlier moment.

The reported surplus cannot be used as proof that the April increase was unnecessary or excessive. It included grants, and the package does not close timing, cash needs, capital expenditure, reserves, invoice cycles or the category mix. Conversely, general startup language cannot prove that USD 6,500 was necessary. A need for recurring finance does not derive a particular price without the expected number of payers, the budget gap, the allocation rule, cash-flow assumptions and alternatives.

Scale illustrations help show why the missing population matters. One full USD 6,500 annual Medium fee would equal about 1.4213% of the reported aggregate 2005 membership-fee income; the USD 1,500 increment would equal about 0.3280%. These are denominator comparisons only. They are not evidence that a full annual invoice applied in 2005 or that any identified member paid. If M is the unresolved number of full annual Medium billings, the nominal revenue difference is USD 1,500 multiplied by M. Because M is unknown, actual Medium revenue and its share of AFRINIC income cannot be calculated.

One table in the meeting report creates a further caution. It states income of USD 491,000, operating expenditure of USD 360,908 and capital expenditure of USD 66,400, leaving arithmetic headroom of USD 63,692. Yet the table is titled “Budget for 2006/2007” while report highlights refer to a 2005 budget. Without a corrected report or underlying approved budget, USD 491,000 is not a clean 2005 necessity denominator. Apparent numerical precision cannot cure a disputed year label.

The defensible financial finding is therefore narrow. AFRINIC documented a startup objective of replacing incubation support with a fee-funded model, and its later accounts confirm that fees were an important source of aggregate income. The public record does not close the financial necessity or derivation of precisely USD 6,500 for Medium. That amount could have been pragmatic, rounded, negotiated or modelled; the missing Board paper and forecast prevent choosing among those possibilities.

Price, renewal and the shape of the middle

The Medium change was the smallest of the three upward amendments recorded in the same fee block, both in dollars and percentage. Medium rose USD 1,500, or 30%. The recorded Large figure rose by USD 6,000, or about 85.7143%, and the Extra Large figure rose by USD 5,000, or about 33.3333%. Those sibling figures are relevant only to locate Medium on the altered fee curve. They do not justify the Medium amount and are not a basis for importing the different incidence and concentration questions that belong to those categories.

The Medium-to-Large nominal gap widened from USD 2,000 to USD 6,500, while the Large-to-Medium ratio moved from 1.4 to exactly 2.0. Using the USD 2,250 Small figure printed in both checked schedules, the Medium-to-Small ratio moved from about 2.2222 to about 2.8889. This is evidence that the shape of the schedule changed around Medium even though Medium received the least severe upward amendment. It is not evidence that the population within any row stayed constant, or that cost rose in those proportions.

For a Medium member comparing standard annual charges, the increase can be expressed as USD 125 per month if the annual difference is divided by twelve. That is a scale translation, not evidence of monthly billing. Under the stated early-renewal condition, the printed discounted amount moved from USD 4,500 in Draft01a to USD 5,850 on the post-decision page, a USD 1,350 or 30% increase. The apparent saving from paying under the discount condition rose from USD 500 to USD 650.

This renewal structure could affect behaviour in more than one direction. It could reward early payment and improve AFRINIC’s cash predictability. It could also make the timing of payment more consequential for an operator facing a recurring fixed charge. But there is no eligibility or payment ledger. We cannot say how many members renewed early, whether the discount altered arrears, or whether operators treated it as an incentive rather than a routine condition.

Predictability itself has two dimensions. A published band and stated renewal price can let an operator forecast the official charge. Yet the 30% revision, unresolved effective date and missing transition rule leave uncertainty about the path from approval to the first affected renewal. The checked record tells a member what the later schedule printed, but not what notice each member received or how an existing annual cycle was handled.

Who bears the increase—and what remains unseen

At the first level of incidence, the contractual cash burden falls on an LIR classified as Medium under the applicable schedule. The nominal standard difference is USD 1,500 annually. That statement does not identify any payer. No sealed evidence gives a Medium member’s name, revenue, margin, exchange-rate exposure or ability to pay.

At a second level, an operator could absorb the charge, pass some portion to customers or alter expansion and renewal decisions. These are plausible economic mechanisms, not recorded historical outcomes. There are no customer counts, tariff changes, pass-through rates, investment records or affordability measures. The article therefore cannot say that end users paid more, that deployment was delayed, or that an operator was harmed.

At a third level, the registry may receive more predictable recurring finance, helping maintain accurate coordination records and member-facing services. Yet actual contribution again depends on the unresolved number of Medium payers, invoice timing, discount use and collection. One cannot infer resilience from a multiplication whose central variable is missing.

The incidence question also exposes the limitation of billing labels. “Medium” is an administrative class associated with a published holdings band. It does not prove a member’s economic scale, political standing or service use. A /18 endpoint and a /16 endpoint have different nominal per-address ratios under the same fee, but the record does not tell us which has more customers, more revenue or more staff demands. Holdings are evidence of the schedule’s chosen proxy, not a complete theory of capacity.

Authority without invented sovereignty

The authority ledger for this event is modest. The AFRINIC Board is the recorded maker of the four amendments. The AGM is the recorded approver of the figures. AFRINIC’s registry administration applied the published classification row within its private service and membership relationship. An admitted LIR could receive services subject to the agreement, payment and resource-policy criteria; payment itself did not guarantee an allocation.

Several larger propositions remain unsupported. The record does not establish that paying a higher fee or holding more resources bought more formal votes. A 2007 bylaw with a one-member-one-vote clause is later evidence and cannot be projected backward to prove the rule governing the 2005 approval. The record also does not establish the absence of informal influence. Both the formal voting mechanism and any category-based bargaining effect remain open.

More fundamentally, the fee was a private membership and registry charge, not a tax imposed by a sovereign. AFRINIC coordinated a ledger of unique number resources and provided related services. The organisation did not create the economic value in the networks, equipment, customers and operational work that used those numbers. Approval of a billing schedule did not grant AFRINIC ownership of an operator’s resources or infrastructure, and it did not turn the AGM into a continental legislature.

Nonpayment could matter within the demonstrated contractual and registry framework. It could affect contractual services or recorded membership status under applicable rules, and AFRINIC could pursue lawful private remedies it actually held. It could not, by inherent sovereign power, imprison an operator, impose a state fine, physically disable routers or erase the fact that a network continued to operate in the real world. The ledger has operational importance, but it remains a ledger; coordination is not rule over the underlying infrastructure.

This doctrine boundary does not make the charge voluntary in every practical sense, and it does not extinguish a valid private obligation. It keeps the source of power accurate. A member may rely on registry services and face contractual consequences for nonpayment without AFRINIC acquiring the penal authority of a state. Likewise, recorded AGM approval may support an internal corporate decision without constituting consent from every African network, user or government.

The evidence-weighted verdict

On holdings, the result is affirmative but bounded. AFRINIC’s contemporaneous records connect the LIR classes to allocated or held number resources, and the published Medium row covers /18 to /16. The USD 6,500 charge therefore redistributed common registry finance through a holdings band. Because one fee spanned a fourfold single-prefix address range, it was a stepwise proxy rather than an exactly proportional address charge.

On category-specific service burden, the result is not closed. AFRINIC documented a common service surface and payment-before-service sequence. It did not disclose Medium workload, marginal cost or a driver linking the extra USD 1,500 to support, training, allocations or database operations. There is no evidentiary basis for saying that service burden rose 30% or that Medium caused a particular share of cost.

On financial necessity, context is supported but the amount is not justified by the available public record. Startup self-finance was a real recorded objective, fees were a material income source, and dependence on transitional support formed part of the institutional setting. The missing Board model, payer count, cash-flow forecast, per-tier collections and alternatives mean that the necessity of precisely USD 6,500 remains unresolved. Ex-post aggregate surplus neither proves nor disproves what was necessary ex ante.

The Medium increase is consequently best understood as a documented reallocation through a holdings-banded schedule, undertaken in a startup-finance setting, with its service-cost causation and exact amount derivation left open. That conclusion is less sweeping than a verdict of fairness or excess. It is also more useful: it identifies the mechanism the evidence actually shows and the records needed to test the rationales it does not.