In Cape Town, AfriNIC seeks to safeguard its community governance

A warning in the inbox

On 21 July 2026, a message to a mailing list devoted to Africa’s Internet number resources exposed a growing unease. Hytham El-Nakhal, a co-chair of AfriNIC’s policy working group, reported an accumulation of emails repeating the same objections. The discussion concerned a proposal on AS-SET names, objects used to describe groups of networks. Repetition was burdening participants, the chairs and registry staff. The co-chair reminded the group that multiplying identical objections did not change the assessment of consensus.

The episode could easily be dismissed as a quarrel among specialists. Yet it concerned the mechanism through which part of the rulebook for Africa’s Internet is written. In this setting, a mailing list is neither a simple information channel nor a comment section attached to decisions made elsewhere. It is a working space where proposals are presented, challenged, revised and preserved in a public record. When that space becomes difficult to follow, the community’s ability to produce understandable and broadly acceptable rules is weakened.

As AfriNIC-38 approaches, scheduled for 18 November at the Hyatt Regency in Cape Town, that tension runs through the agenda. The meeting is open to everyone, in person and remotely. The proposals address the organisation of the working group, IPv4 exhaustion, IPv6 deployment, policy compliance and the operation of the mailing list. Their common thread is less visible than their technical titles. They concern the practical conditions under which a community can continue to govern its resources.

Why Africa established its own registry

AfriNIC, the African Network Information Centre, is the regional Internet registry serving Africa and part of the Indian Ocean. Its institutional history traces the initiative to an INET workshop in Malaysia in 1997. Incorporated in Mauritius in 2004, it was recognised by ICANN in April 2005 as the 5th regional registry. The initiative sought to give African circumstances a distinct place in the administration of Internet number resources.

The global Internet number registry system coordinates the uniqueness of these resources through IANA, regional registries and the organisations receiving them. AfriNIC belongs to that system alongside APNIC, ARIN, LACNIC and the RIPE NCC. It is neither a continental ministry of digital affairs nor the owner of everything carried over African networks. Its particular responsibility concerns resources that allow networks to identify themselves and operate together.

An IP address identifies an interface on a network using the Internet Protocol. Addresses are grouped into blocks distributed under shared rules. An autonomous system number, or ASN, identifies a collection of networks under a coherent technical administration. An Internet service provider, a large company or a university network may need these resources to organise its connectivity. Their distribution therefore affects very practical opportunities for growth and interconnection.

The registry’s political significance lies in that connection between infrastructure and rules. Cables, data centres and antennas cannot by themselves produce digital autonomy if the actors using them have no influence over access to the resources necessary for their operation. Establishing an African registry brought part of that rule-making capacity closer to the continent. It did not make the capacity automatic, irreversible or independent of the people willing to exercise it.

The board, the members and the community are different constituencies

Understanding the disputes requires separating several functions. AfriNIC’s staff administer services and implement policies. Its board carries responsibility for the organisation’s governance. Members have rights defined in its institutional documents. The Policy Development Working Group, or PDWG, brings together participants in the policy-making process. These groups overlap, but they are not interchangeable.

Section 3 of the Consolidated Policy Manual provides for open participation, online or in person. Work takes place on the Resource Policy Discussion mailing list, known as RPD, and at public policy meetings. Under the currently published process, the co-chairs assess consensus, a last-call period allows final scrutiny, and proposals that complete the required stages are submitted to the board for ratification. An appeals procedure is available.

This architecture prevents policy-making from becoming solely a negotiation among organisations that already hold resources. A researcher, student, association or new operator can raise a useful objection without owning a large address block. Expertise can come from operating a network, but also from identifying an exclusionary effect, a disproportionate cost or an ambiguity that will affect future entrants.

Open participation, however, does not remove responsibility for what a contributor says. Participants should be able to explain their reasoning, correct mistakes and clarify the interests they represent when these help others understand their interventions. The challenge is to make that accountability workable without turning an open institution into a circle restricted to people already recognised by its longest-standing participants.

The distinction between a resource holder and a member of the organisation is central to our earlier article on Cloud Innovation’s appearance in the register of members. Rights under an allocation agreement, entry in the members’ register and participation in policy development raise different questions that should be examined separately.

The mailing list is part of the decision-making infrastructure

The public RPD archives preserve exchanges in their sequence and their contradictions. They allow readers to recover an earlier objection, examine an author’s response or compare successive versions of a proposal. This traceability matters when a rule eventually imposes obligations on organisations whose representatives were not all present in the meeting room.

Email also offers an accessibility advantage. An engineer dealing with an outage, an academic far from the venue or an operator with a limited travel budget can contribute at another time. A written argument can be reread, translated and tested against local experience. Public meetings provide another form of engagement, but they should not erase the work carried out between meetings.

That openness depends on a resource less visible than bandwidth. Reading, comparing and responding require attention. Producing 1 message may take only a few minutes. Checking its claims, reconstructing the exchanges to which it refers and deciding whether it contains a new objection can take much longer. When this imbalance grows, participants with the least time are among those most likely to withdraw.

The problem is therefore not simply the number of emails. A busy discussion can be productive. The concern arises when readers cannot readily distinguish new arguments from repetition, connect replies to the right proposal or understand what remains genuinely unresolved. A large archive is not necessarily a usable institutional memory. Its readability is part of the quality of governance.

Consensus is not an email count

The term rough consensus can be misleading. It does not authorise careless decision-making. RFC 7282, published in 2014, explains an IETF practice centred on addressing objections rather than counting them. Unanimity is unnecessary, but concerns must be considered. The document is informational. It illuminates the discussion without replacing AfriNIC’s own process.

Consider a hypothetical example. 40 participants support a proposal, while 1 demonstrates that it would prevent a category of networks from meeting an essential requirement. The number of supporters does not fix that defect. Conversely, 40 messages reproducing an objection that has already been examined and adequately resolved do not automatically create 40 separate problems. The task is to establish what remains unanswered, precisely enough for others to examine that assessment.

This judgement places a responsibility on the co-chairs. A credible conclusion should make clear which objections were accepted, which were resolved and why some no longer prevent the proposal from moving forward. Declaring consensus is not enough. Participants, especially those who remain unconvinced, should be able to understand the reasoning that led to the conclusion.

The classification of an objection as technical also requires care. Implementation costs, access to necessary data, effects on a small operator and the ability to correct a decision are not detached from how networks function. A broad concern must be made specific and supported with evidence. But a demonstrated difficulty does not become irrelevant merely because it has an economic or institutional dimension.

Ben Roberts warns of an artificial appearance of community support

On 18 July, Ben Roberts announced on RPD an article about astroturfing, the manufacture of activity designed to resemble a spontaneous grassroots movement. In that announcement and a 24 July message, he described contributions he considered repetitive, artificially produced and capable of manufacturing an impression of consensus. These are his allegations and interpretation of the exchanges, not the findings of an independent investigation presented here.

His warning raises a question that extends beyond any particular software. If an open process in practice rewards apparent presence, there is an incentive to multiply the signs of participation. Messages may appear to express a diversity of voices while contributing only a single line of argument. That possibility deserves an institutional response. It does not establish coordination, a fictitious identity or a sponsor behind particular individuals without additional evidence.

The 23 July thread also records objections to this interpretation. Several participants ask for suspicions to be assessed against evidence and neutral rules. They distinguish writing assistance, similarities between arguments and conduct that is actually deceptive. They also point out that professional affiliation does not automatically mean every contribution is made on an employer’s behalf.

Those responses belong in the record. Dismissing them solely because they challenge the allegations would make the allegations impossible to contest. A community can recognise a problem of saturation while refusing to presume that every newcomer is a fabricated participant. The distinction should rest on documented behaviour, disclosed interests and the quality of arguments, rather than a name, nationality or writing style.

AI tools make the difficulty more acute. They can help someone translate a contribution or express an objection clearly in a language they do not speak fluently. They can also produce large amounts of text that nobody is genuinely prepared to defend. The relevant requirement remains the intellectual responsibility of the person submitting the message. Can that person explain the argument, answer a specific question and acknowledge when an objection has been addressed?

An IP address cannot establish a person’s identity on its own

Technical traces can assist an inquiry into the authenticity of participation. They must be interpreted according to what they actually show. Email headers may contain information about the servers through which a message passed. The SMTP specification describes the addition of transit information. A mail relay’s address is not necessarily the address of the author’s device, much less proof of nationality or university enrolment.

Similarly, a DKIM signature associates a signature with a domain and supports certain checks on a message. By itself, it does not certify the civil identity of whoever composed the text. Moving from a technical indicator to an allegation of impersonation requires a stronger chain of verification than a location displayed on a website.

A serious procedure should specify what fact is being established, who is authorised to examine the data, what explanation the person concerned can provide and what information must remain confidential. Claiming to protect the community while unnecessarily exposing participants’ personal data would be contradictory. Verification should rebuild trust and allow a response, rather than fuel an open-ended collective investigation without clear responsibility or limits.

The danger of treating the community as terra nullius

The expression terra nullius can serve here as a political metaphor, provided it is not mistaken for a legal description. An open mailing list is not vacant territory. It is the product of accumulated work, relationships of trust, rules and obligations. Treating it as space available to whoever can occupy it most intensively erases that collective construction.

The danger can develop gradually. Regular contributors tire of reformulating the same responses. Newcomers lose their way in overlapping threads. Operators absorbed in urgent work postpone their contributions. The forum remains formally open, but the cost of taking part becomes too high for some of the people it is intended to include. Participation may then concentrate among individuals or organisations able to devote much more time to it.

This interpretation does not demonstrate that a particular actor organised the dispossession of RPD. It explains why exhaustion of attention is a governance issue even where no coordinated operation has been established. The troubling outcome would be a community repeatedly invoked in speeches but increasingly unable to deliberate on equitable terms.

Community sovereignty therefore cannot be defended only against external intervention. It also depends on the responsibilities African actors are willing to assume. Obtaining resources and then treating the rules that make them usable as somebody else’s work weakens the shared system. This is an argument for participation, not a general accusation of indifference against the continent’s operators.

Redistributing authority without moving the problem elsewhere

The PDP WG Guidelines and Procedures proposal, by Grégoire Ehoumi, Noah Maina and Adeola A. P. Aina, envisages 3 chairs with staggered terms and a Regional Number Council combining them with the 3 NRO representatives. It also provides an exceptional mechanism when board ratification does not occur, involving a petition supported by 15 contacts from distinct members across at least 3 subregions. Institutional amendments would be necessary. These are proposed arrangements, not rules already in force.

Continuity is the central issue. If an essential stage depends on a body that cannot function, the entire policy process may become immobilised. In an analysis published on 6 December 2025, Alain Patrick Aina examines weaknesses exposed by the registry’s crisis and argues for a review of the framework. His account helps explain the search for mechanisms that can sustain community work when institutional arrangements break down.

Creating additional bodies or a substitute route does not, however, guarantee independence. Participants need to examine incompatibilities between roles, recusal procedures, publication of reasons and the conditions for appeal. A person assessing a dispute should not end up judging their own conduct. The separation of functions therefore deserves as much attention as the number of seats.

The scope of authority matters equally. Adopting a resource policy does not by itself amend every bylaw governing the organisation. The community needs to know which provisions could take effect directly, which depend on other decisions and what arrangements would apply during transition. Protection against paralysis should not create new uncertainty about who has the authority to act.

Making the meaning of a decision more explicit

Through Update of PDP, Jordi Palet Martinez proposes clearer provisions on consensus, the combined role of mailing-list and meeting discussions, timing and impact assessment. The text would permit formal identity verification with regard for data protection. It requires impact analysis within 4 weeks and at least 1 week before a meeting, or a justified draft where a complete analysis is unavailable. These provisions remain under discussion.

Impact analysis is essential to keeping the debate grounded. It should help participants understand what the registry would need to develop, what members would have to declare, how much additional work would result and what implementation difficulties can reasonably be anticipated. A proposal can pursue a legitimate objective while relying on an instrument that is too costly, imprecise or difficult to administer. Making those trade-offs visible strengthens deliberation rather than merely delaying it.

Identity verification requires a distinct assessment. Establishing that a real person controls an account is different from requiring membership of a profession or possession of resources. The circumstances triggering verification should be defined, the data collected limited and the administration impartial. A rule intended to discourage multiple identities would lose legitimacy if it became a selective means of excluding opponents.

The 2 proposals reforming the process also need to be read together. They may complement each other in some areas and create difficulties of interaction in others. A useful test is to follow a concrete case from submission through any appeal, identifying who intervenes, within what deadline and under whose scrutiny. That practical journey is how a procedure should be judged.

Linking scarce IPv4 resources to IPv6 deployment

IPv4 uses 32-bit addresses, allowing approximately 4.3 billion combinations before reserved uses are taken into account. IPv6 expands addresses to 128 bits and vastly increases the available space. That difference explains the strategic importance of transition. It does not instantly remove the equipment, services and economic constraints of networks still dependent on IPv4.

The IPv6 as Criteria in IPv4 Soft Landing proposal, by Jordi Palet Martinez and Jaco Kroon, links future IPv4 requests to IPv6 progress. It establishes thresholds of 25%, 50%, 75% and 95% at 12, 24, 36 and 48 months, subject to the specified starting conditions. Metrics differ for access networks and content or hosting activities. Applicants without IPv6 resources would also request them alongside IPv4.

The principle is to use a scarce resource as an incentive to prepare for the future. Assessing it requires scrutiny of what the thresholds measure. Making a connection capable of using IPv6 does not guarantee that all its traffic will use that protocol. For a hosting provider, service availability raises different questions from activating thousands of access lines. The distinction is essential to avoiding misleading comparisons.

A small provider should be able to explain constraints involving equipment, upstream connectivity, training or finance. That does not justify an indefinite exemption, but it helps establish whether proposed milestones are achievable. The discussion should address support, acceptable evidence and the treatment of temporary difficulties. An obligation that accelerates deployment at large networks while closing access to new entrants would have distributional effects that deserve explicit consideration.

The issue is therefore not a choice between IPv4 and IPv6 as opposing camps. It is how to organise their coexistence during transition and distribute the effort fairly. Deploying IPv6 and accounting for the use of IPv4 resources are complementary objectives. Progress on one does not remove responsibility for the other.

Monitoring compliance without turning every signal into a sanction

The Policy Compliance Dashboard proposal, by Jordi Palet Martinez, Frank Habicht and Mark Elkins, envisages a dashboard within MyAfriNIC. Automated checks would help members and staff identify potential compliance issues. The text distinguishes alerts from staff verification and does not make every signal an automatic decision to revoke resources. Implementation is envisaged in phases.

Such a tool could make obligations more visible before a dispute escalates. An unreachable administrative contact, an outdated registration entry and a substantive policy violation are nevertheless different situations. The dashboard should explain where an alert comes from, which rule it concerns, what information needs correction and how a member can challenge a mistake.

In an 8 October exchange, Andrew Alston raises concerns about the network information that might be requested and its confidentiality. This objection deserves a concrete response. What data are strictly necessary? Who will have access? How long will they be retained? Transparent criteria do not require publication of every member’s commercial or technical configuration.

The distinction between data and conclusions is particularly important for IP resources. Routing or geolocation information can assist an investigation without establishing a contractual violation on its own. AfroDIG’s previous investigation into IPinfo measurements emphasised the need to separate technical observations from their interpretation. The same discipline should guide a compliance tool.

Credible oversight should protect the registry against abuse while protecting members against errors in oversight. These objectives reinforce each other. A reasoned decision, supported by verifiable evidence and an opportunity to respond, will be stronger than a sanction whose main justification is the colour displayed on a dashboard.

Managing scarcity without reserving the future for established operators

In an article dated 14 October 2025 on IPv4 allocations per capita, Ben Roberts had already highlighted differences between African countries. That approach raises a useful distributional question, but its indicator needs careful interpretation. An address registered against a country does not necessarily correspond to equipment located there or to a person who is actually connected. Allocation, resource use and population access are 3 related dimensions that require different measurements.

The Dynamic IPv4 Pools Exhaustion Management Framework, proposed by Aina, organises available or recovered resources into 4 pools according to origin. After exhaustion, it provides for a waiting list, bounded allocation sizes and limits on requests associated with the same organisation. Temporary transfer restrictions seek to reduce opportunities for rapid turnover of newly obtained resources.

For a non-specialist reader, notation such as /24 or /23 indicates the size of an IPv4 block. A /24 contains 256 addresses and a /23 contains 512. As the number after the slash increases, the block becomes smaller. Setting a minimum or maximum size is therefore more than an administrative convention. It affects the practical usefulness of an allocation and the number of organisations that can receive resources.

The debate needs to distinguish growth requirements at established operators from entry conditions for organisations that do not yet exist. Consistently favouring demonstrated historical utilisation can benefit the largest networks. Conversely, a ceiling set too low can prevent a viable project from starting. No scarcity rule is neutral in its effects, even when expressed in the apparently objective language of prefixes and percentages.

It is also necessary to distinguish resources that might theoretically be recovered from resources actually available for redistribution. Verification, applicable procedures and possible disputes can separate those situations. A waiting mechanism should make the order of processing, the status of requests and relationships between organisations understandable. Trust will depend as much on this daily administration as on the general formula adopted in Cape Town.

Lu Heng and Cloud Innovation, why more than 6 million addresses should return to AfriNIC

A debate about scarcity cannot leave aside the resources whose return AfriNIC has sought. In its 27 August 2022 communiqué, the registry stated that it had allocated approximately 6.2 million IPv4 addresses to Cloud Innovation, a company controlled by Lu Heng, on the basis of needs expressed for the region and a registration services agreement. This figure describes the allocations involved in the dispute. It is not a pool currently available for distribution to new applicants.

On 8 July 2021, the board announced their reclamation, citing contractual breaches and allowing users 90 days to migrate. Cloud Innovation disputes that action and relies on protections obtained through the courts. The interlocutory judgment of 19 July 2022 records measures preventing termination pending resolution of the dispute and rejects the abuse-of-process objection then raised by AfriNIC. Those decisions must be distinguished from a final judgment on whether all the allocations complied with the applicable requirements.

Criticism of their use does not rest solely on the African registry’s position. In a 27 August 2021 statement, ARIN president John Curran assessed that most of these addresses were being used outside Africa and supported their recovery for the African community’s benefit. He also referred the contractual issues and questions of possible fraud to the Mauritian courts. This was another registry’s public assessment. Its significance should not be converted into a judicial verdict.

The position advanced here is clear. Returning these resources to AfriNIC’s pool, through lawful procedures and on the basis of established breaches, should be a strategic priority. Defending community governance cannot stop at preserving an institution while resources remain beyond the redistribution it is entrusted to organise. The objective should be to restore their usefulness to African networks and African needs. That requires a sound evidential record, an opportunity for both sides to be heard and implementation consistent with court decisions.

Recovery should operate within soft landing, the framework governing the distribution of remaining IPv4 resources as they run out. Soft Landing, Recovered Space and Priority, ratified on 25 September 2026, provides for returned or recovered resources to enter the available pool after a 12-month quarantine, which may be shortened for operational reasons. Ratification and implementation are separate stages. This policy determines no individual dispute and does not, by itself, make Cloud Innovation’s addresses available.

The issue also connects with digital public infrastructure, or DPI. UNDP describes DPI through systems including identity, payments and data exchange. IP addresses do not, on their own, constitute digital public infrastructure. They contribute to the connectivity and hosting on which those services depend. Returning resources to circulation could therefore support local operators, community networks, research institutions and hosting providers that make these services accessible, subject to the applicable eligibility requirements.

This is neither a promise of 6.2 million newly connected people nor a case for arbitrarily reserving blocks for government programmes. An address can serve multiple users, and allocation does not replace fibre, electricity or operational expertise. The community should establish transparent rules proportionate to demonstrated needs, with verification and an appeals process. Recovery should widen access to resources, prevent renewed concentration and support IPv6 deployment, so that any breathing space advances the transition rather than prolonging dependence.

For the board, this priority should mean a documented recovery strategy, identified legal and technical resources, and public progress reporting consistent with confidentiality obligations. For the courts hearing the cases, the issue is to determine the substantive questions and enable effective decisions to be implemented, with full respect for judicial independence. For the community, the task is to prepare redistribution arrangements, demand accountability and participate in policies that prevent renewed concentration. These responsibilities reinforce one another without becoming interchangeable.

The address dispute must also be distinguished from the winding-up petition previously examined by AfroDIG. Preserving the registry’s continuity and recovering disputed resources are connected but legally distinct objectives. The community should be able to follow progress on each. The outcome to seek is more than a procedural victory. It is restored capacity to distribute resources to African networks and their future users.

Meeting dates also determine who can take part

Jordi Palet Martinez’s Public Policy Meeting Dates proposal seeks to avoid clashes with previously announced meetings of other organisations, including RIRs, the IETF, ICANN and the Internet Society. It calls for dates to be published at least 1 year ahead, subject to specified exceptions. What looks like a scheduling issue is also a question of access.

Someone who must apply for funding, obtain a visa or arrange time away from work does not enjoy the same flexibility as an organisation with a permanent team. Predictability can broaden participation. Conflicting dates may force experienced contributors to choose between responsibilities that would otherwise reinforce one another.

The counterargument also deserves attention. In exchanges on 28 August, Ben Roberts highlights the economic value of bringing some events together to reduce travel. For an African operator, attending several meetings during 1 trip may be more affordable than financing separate stays. Avoiding every proximity in the calendar therefore does not, on its own, guarantee better participation.

A sound assessment should distinguish a clash that forces someone to miss a session from scheduling that makes travel more worthwhile. It should also consider remote-participation hours, document availability and the opportunity to submit a contribution before the meeting. Hybrid participation means more than providing a video link.

Repairing the conversation includes repairing its tools

Through RPD Mailing List Setup and Guidelines, Benson Muite proposes presentation and operating rules intended to improve readability. The proposal addresses message format, threading and replies to compilations known as digests. It would disable digest mode unless average traffic exceeded 20 messages a day over 2 months.

Subscribers know the problem. A reply to a compilation covering several discussions can carry a generic subject line and include lengthy exchanges unrelated to the objection being raised. The archive preserves the message, but readers must reconstruct its context. That burden is repeated for each participant and then for the co-chairs when they prepare a summary.

In a 14 October message, Ben Roberts returns to this difficulty. Reminders about courtesy are not the only possible response. Software choices, default settings, clear rejection notices and guidance for new subscribers can help prevent mistakes instead of merely sanctioning them afterwards.

The limits of the reform should nevertheless remain clear. A better organised list does not prove its authors’ identities, reveal possible sponsors or resolve substantive disagreement. It improves the conditions for collective work. That modest gain is precisely its value, provided technical requirements do not exclude people using less flexible email tools or limited connections.

What the community should prepare before 18 November

The draft programme also includes policy implementation experience, a PDWG chair report, an open microphone, a working-group chair selection and an exchange between the board and the community. These sessions matter alongside the proposal presentations. They provide opportunities to ask what actually happened after decisions were made and what difficulties the texts failed to anticipate.

Useful preparation begins with reading the specific version of each proposal. Version numbers matter because an objection may already have been addressed in a later draft. Participants can then identify the obligation that would affect their organisation, work through an implementation example and request clarification where the outcome remains uncertain. A documented case often contributes more than a general statement of support or opposition.

Operators can bring practical constraints into the discussion without disclosing sensitive commercial information. Universities can examine whether criteria are coherent. Associations can consider effects on new entrants and whether participation is genuinely accessible. Public authorities can follow the public-interest implications without turning every technical disagreement into a diplomatic confrontation. This diversity becomes useful when it generates questions that can be examined against evidence.

For newcomers, the RPD subscription information, policy manual and archives offer a starting point. Contributions should stay within the relevant thread, use a precise subject and distinguish observed experience from a hypothesis. Saying that a provision might create a problem is a first step. Explaining for whom, under what circumstances and with what possible correction helps the discussion move forward.

Work will continue after the meeting through summaries, formal procedural stages and implementation. Community attention should not end when the video call closes. An adopted rule may still reveal implementation difficulties that need to be documented and, where necessary, corrected. Revision is part of responsible governance.

Sovereignty also depends on taking the time to read the rules

AfriNIC cannot safeguard its governance simply by invoking its community character. That character must remain observable in how arguments are considered, responsibilities distributed, conflicts of interest handled and decisions explained. An open process becomes fragile when saturation displaces deliberation. It is equally fragile if defending it means restricting participation to people who already know one another.

The Cape Town meeting offers an opportunity to work through that tension without reducing it to a contest between individuals. The 7 proposals do not all share the same diagnosis and do not deserve blanket endorsement on principle. They should be assessed separately and in combination, asking whether they make the institution more understandable, accessible and capable of acting.

The registry was created so that African circumstances could carry weight within a global system. Preserving that achievement requires more than symbolic attendance at conferences. It means reading a draft, defending an objection, recognising a convincing answer and returning to check implementation. The community also exists in that ordinary work. Its capacity to sustain it is what Cape Town will put to the test.