Summary

  • AFRINIC ratified AFPUB-2020-GEN-006-DRAFT03 on 4 February 2026 with its decisive asymmetry intact: an AFRINIC member may transfer a “Global” resource to another RIR region, and a Legacy holder has a separate outward route, but an ordinary AFRINIC-pool resource classed “Regional” cannot use that outbound path.
  • AFRINIC says the Regional label does not alter WHOIS visibility, routing rights or day-to-day use. That clarification is important, but it does not make the restriction trivial: it confirms that geography acts at the transfer and registry-recognition layer, where it can narrow exit, liquidity, collateral value and bargaining power.
  • The policy passed last call and a co-chair consensus decision in January 2022. A subsequent appeal was rejected on admissibility because the required prior attempt at resolution was not shown; the Appeal Committee expressly did not decide the substantive objections.
  • Ratification came while the court-appointed receiver had not been formally discharged and litigation was challenging the elected directors. Skyconnect filed a direct challenge to the ratification 33 days later. The public record available by 10 August 2026 contained no final merits ruling either validating or invalidating the policy.
  • Ratification did not itself create a working inter-RIR service. ARIN described implementation as pending with a date to be determined, its live table still marked AFRINIC as not approved, and RIPE NCC said transfers to or from AFRINIC were not available at the evidence cutoff.

On 4 February 2026, AFRINIC did more than announce a future way for number resources to move between registries. It ratified a rule that asks where an address came from before deciding where its holder may send it next. Under D3, the Number Resources Transfer Policy, an IPv4 block received from another RIR is “Global” and may, on paper, leave the AFRINIC service region again. A pre-RIR Legacy resource also has an outward route. But an ordinary resource issued from the AFRINIC pool is “Regional”, and that category cannot pass through the ordinary member-to-other-region transfer channel.

A fourth category, Reserved, is generally closed to transfer except in a merger, acquisition or takeover.

The conflict is therefore unusually concrete. It is not about whether addresses can be routed worldwide; AFRINIC says the classification does not change routing or daily use. It is about whether the registry will recognise a holder’s exit transaction. AFRINIC presents the system as stewardship of scarce resources, accurate registration and controlled redistribution. NRS, Heng Lu, LARUS and BTW identify the opposite institutional mechanism: a service boundary is being converted into an asset boundary by an organisation whose legitimate function is bookkeeping, uniqueness and coordination, not sovereign control of capital.

The stakes extend beyond one policy label. A narrower buyer set can reduce optionality, weaken price discovery, complicate financing and discourage resources from entering a regime perceived as difficult to exit. Yet the policy’s legal and practical condition is also unsettled. It was ratified by a Board whose authority was under challenge while a receiver remained in place; a court case then directly contested ratification; and, 187 days after the announcement, the public operational evidence still showed no live inter-RIR route.

To understand D3 properly, the text, the procedure, the corporate authority question and the implementation state must be kept separate. Collapsing them produces either an official success story or a sweeping legal accusation. Neither is accurate enough.

Four labels, two ordinary ways out

D3’s architecture begins with origin. Resources obtained from IANA, AFRINIC allocation or recovery, or the Early Registration Transfer project are placed in the Regional category. Special-purpose holdings, including specified critical-infrastructure resources and resources associated with the exhaustion phase of soft landing, are Reserved. Resources issued before the RIR system and marked as such by AFRINIC are Legacy. Resources brought into AFRINIC from another RIR through an inter-RIR transfer are Global.

The labels then determine mobility. Sections 3.3.3 and 3.3.4 carry the crucial rules. A Legacy holder may transfer a Legacy resource to a recipient in another RIR service region. An AFRINIC member may transfer to such a recipient only when the resource is Global. Regional resources may be transferred between AFRINIC members but not through that ordinary outward path. Reserved resources are generally denied transfer regardless of the source and recipient, with the stated exception for mergers, acquisitions and takeovers.

That means only two of the four categories have a non-corporate-reorganisation path out: Legacy and Global. This is a category count, not a measure of affected address volume. No public inventory in the reviewed record showed how many addresses sit in each class, and it would be wrong to turn two out of four into a claim that half the address space is mobile. The asymmetry is nonetheless exact. It is written into eligibility rather than inferred from speeches about a feared resource drain.

D3 also covers autonomous system numbers, while the public ratification overview makes its clearest outward-transfer statement about IPv4. For IPv4, the minimum transfer size is a /24, or 256 addresses. How counterpart registries will implement the ASN categories in every case remains a live question. The policy also requires the source to establish rightful control, requires recipient eligibility, addresses disputed resources, and makes transfer dependent on the other RIR’s compatible process. Those are real coordination needs. They do not erase the origin-based wall.

The policy reserves express written approval to AFRINIC and says transfers outside the approved policy will not be recognised. Proof of control, protection against duplicate claims and accurate recipient records are proper registry tasks. The authority problem arises when the same approval surface decides the economic geography of a lawful transaction. A bookkeeper may decline to record a fraudulent or conflicting instruction. It does not follow that the bookkeeper owns the transaction or can legitimately restrict its destination because of the administrative history of the asset.

AFRINIC’s own clarification helps isolate the issue. It says Regional is an administrative classification, not a visible WHOIS label, not a change to routing rights and not a restriction on day-to-day use. Existing allocations and valid registration-service agreements were not described as revoked merely by ratification. This is a useful limit on overstatement: D3 did not switch off networks on 4 February. But the clarification also identifies what the category actually does. It governs transfer eligibility. Packets remain globally routable while registry-recognised alienability becomes regional.

That difference is the core of the policy, not a semantic detail. An operational asset need not be physically immobilised to be economically locked in. A company can continue routing a block and still lose a choice over whom it can sell to, which market can price it, and which registry can record the next holder. A restriction at the ledger layer is especially consequential because counterparties, lenders, insurers and courts use that ledger as evidence. AFRINIC did not create the underlying network reality. Its refusal to recognise a change can nevertheless impose friction on anyone who needs clean records.

The long path from proposal to ratification

Within AFRINIC’s Policy Development Process (PDP), the immediate history begins before D3. AFRINIC had implemented an intra-region IPv4 transfer policy in February 2018, following a predecessor proposal that confined transfers to the region and required approval of recipient need. That was the baseline the later Number Resources Transfer Policy was meant to supersede. The new proposal’s first draft was submitted on 17 October 2020. It already contained four origin or status categories and the principle that only Legacy and inbound resources could move out.

An external reciprocity assessment circulated ten days later recorded objections from ARIN and APNIC to the asymmetry of the outward categories. RIPE compatibility was more favourable, and LACNIC, under the versions examined, did not impose the same reciprocity demand. This matters because “reciprocal” never meant that every registry had identical prose. It meant each potential counterpart had to decide whether the other side’s rule was compatible enough for its own system and policy.

D2 followed on 8 October 2021. AFRINIC’s public-policy meeting in November discussed reciprocity, category restrictions, systems impact and the breadth of AFRINIC’s approval authority. The authors defended regional retention, while the co-chairs requested editorial clarification. D3 was submitted on 22 November. Its revision history described clarity changes; it did not remove the outbound restriction on Regional resources.

The last-call period ran from 8 December 2021 through 5 January 2022—29 calendar dates if both endpoints are counted, or 28 elapsed days. Objections continued in the Resource Policy Discussion (RPD) public mailing-list record. Participants challenged bureaucracy, approval discretion, asymmetric reciprocity and the possibility that a restrictive exit rule would deter resources from coming into AFRINIC. Those messages establish that the objections were made; they do not by themselves prove their economic magnitude.

On 14 January 2022, co-chairs Vincent Ngundi and Darwin Da Costa confirmed consensus and said the report would go to the Board. Their decision is a procedural fact within the Policy Development Working Group. It is not evidence that every AFRINIC member, resource holder, network, customer, company or state consented to the policy. Rough consensus can guide a coordination process, but it cannot convert a voluntary policy forum into the juridical owner of all resources administered by the registry.

Mathanya Ramaboea appealed with support from Cheken Chetty, Oluwabunmi Egbeyemi and Elvis Ibeanusi. The relevant documents are labelled DRAFT02 even though the last-call sequence concerned the D3 text submitted in November. The reviewed record does not resolve whether that mismatch was clerical, version-based or legally significant, so it should not be silently harmonised.

The Appeal Committee met on admissibility on 13 July 2022. It found no evidence that the appellant had first attempted to resolve the dispute with the co-chairs or the working group, as section 3.5.1 of the Consolidated Policy Manual required. The Committee therefore set the appeal aside as inadmissible. Crucially, it expressly did not decide the merits. “The appeal failed” is too broad a summary. The procedural doorway was closed; the substantive questions about approval power, reciprocity, regional retention and economic impact were left unanswered.

The D3 page records a staff assessment dated 28 February 2022. Staff estimated that implementation would require 12 months and identified extensive changes to MyAFRINIC, hostmaster procedures, resource tagging, agreements, logs and inter-RIR coordination. The legal review raised issues involving legacy contracts, facilitator liability and the treatment of resources brought into the region. An implementation estimate is not an effective date. It is evidence that the work was not a simple wording change.

AFRINIC’s later consolidated financial statements said no policies were implemented during 2022 through 2024 and listed policies pending ratification. Governance interruption offers a plausible explanation for some of the extraordinary delay. Even so, 1,535 days elapsed from D3’s submission to ratification, and 1,482 days from the co-chairs’ consensus confirmation to the Board announcement. Those durations describe the gap; they do not establish which institution or dispute caused each day of it.

A corporate act made under a live authority cloud

AFRINIC’s Bylaws separate the ordinary and emergency paths. Articles 11.2 and 11.3 concern the community policy-development route and Board ratification. Article 11.4 provides a distinct route for urgent Board adoption, with Article 11.5 requiring later community endorsement. AFRINIC’s February 2026 defence invoked the ordinary route. Nothing in the checked record showed D3 as an emergency policy created by the Board under Article 11.4.

In the ordinary design, authors propose, the working group discusses, co-chairs determine rough consensus, an Appeal Committee can review procedure, and a lawfully constituted Board ratifies. That chain can generate a corporate registry policy. It cannot generate sovereignty. Nor is a reference to the Bylaws enough to answer whether the particular Board exercising the power was lawfully and finally constituted.

That question was unusually acute in February 2026. The Supreme Court of Mauritius had appointed Gowtamsingh Dabee as receiver on 12 February 2025. His public description of the role was preservative and governance-reconstituting: hold the ring, preserve the status quo and business value, reconstitute the Board, and support the later appointment of a chief executive. It was not presented as a mandate to legislate permanent control of resource mobility.

A court-supervised election took place in September 2025, and AFRINIC announced eight elected directors on 15 September. The announcement proves that AFRINIC published those results. It does not establish a final judicial answer to every challenge against the directors. The receiver applied for discharge on 8 October, but AFRINIC acknowledged in March 2026 that formal discharge was still pending, judgment was awaited, and proceedings sought to invalidate the elected directors.

AFRINIC attributed D3’s 4 February ratification to that Board. Yet the public announcement and later communiqué did not provide a numbered Board resolution, meeting minutes, the motion, the vote, quorum evidence, recusals, a board paper or a legal opinion explaining the receiver-Board relationship. Absence from the reviewed public material does not prove no internal resolution exists. It does mean outsiders cannot audit the exact corporate act from the announcement alone.

NRS’s first-class position is more direct: the individuals acting as the current Board had not established the lawful authority required to bind members, particularly while the receiver remained and director-validity litigation continued. NRS has organised member protection and represents members who expressly authorise it. That role matters because it puts the authority objection in institutional, not merely individual, form. It does not make NRS the operator of AFRINIC’s registry, policy process, appeals, custody or transfer settlement.

On 9 March 2026, 33 days after ratification, Skyconnect—formerly Skyvision Guinee SA—filed a plaint that AFRINIC’s public court list describes as challenging the Board’s ratification of the transfer policy. A related application for interim relief was declined, and an appeal filed on 7 April remained listed as ongoing. The fact of filing proves a live contest. It does not prove the allegations, reveal the full legal grounds or establish the remedy sought because the complete pleadings were not available in the reviewed record.

The boundary is therefore strict. There was no located final merits judgment declaring D3 valid, and none declaring it invalid. There was no final receiver-discharge ruling in the available record. A refused interim application is not a final victory on the policy’s merits. Any account that calls ratification court-approved or court-nullified moves beyond the evidence.

The authority problem nevertheless cannot be dismissed as background noise. Counterpart registries and commercial parties need to know that the actor approving a transfer system can bind the corporation and maintain the result. A live challenge can alter warranties, delay implementation or make cautious counterparties wait. Those effects are plausible mechanisms rather than measured contract costs, but the underlying uncertainty is documented.

Ratified, compatible and implemented are different states

One of D3’s most revealing features is that its paper status ran ahead of operational reality. AFRINIC ratified the text on 4 February. Fourteen days later it issued a communiqué defending the policy-development process and Board constitution while clarifying the Regional label. Neither act, on its own, rewrote software, trained hostmasters, signed counterpart arrangements or opened a production transfer channel.

AFRINIC’s own policy nomenclature distinguishes ratification from implementation. A numbered proposal can be ratified yet remain pending until it is incorporated into the Consolidated Policy Manual and an adoption or implementation date is announced. The policy-development text contemplates implementation in less than six months after last call unless that expectation is waived. No public waiver, implementation announcement or updated incorporated text resolving D3’s relationship with the older intra-region provision was found in the reviewed material.

Counterpart evidence reinforces that distinction. By April 2026, ARIN staff had moved from the objections recorded in the 2020 assessment to a view that D3 was compatible with ARIN’s needs-based policy. That change shows compatibility to be a versioned institutional judgment rather than a demand for textual symmetry. It also shows why a compatibility statement should not be mistaken for a live route. ARIN said the implementation date was to be determined and activation would wait until AFRINIC staff completed their work.

At the 10 August evidence cutoff, ARIN’s live transfer table still marked AFRINIC as not approved. RIPE NCC’s live inter-RIR page said AFRINIC did not then have an operational inter-RIR policy and that resources could not be transferred to or from it. Current APNIC and LACNIC activation notices specifically naming AFRINIC were not available in the reviewed record. The known matrix was therefore incomplete, but the two live counterpart checks were enough to reject the claim that ratification automatically opened a global market.

No AFRINIC implementation date, transfer form, service-level commitment, fee schedule, production log or completed D3 inter-RIR transfer was located. That supports a precise statement: there were zero publicly proven completions in the material reviewed, not a universal claim that no test, internal development or undisclosed transaction ever happened. Internal work may have been under way. The public operational channel was not demonstrated.

This distinction also changes how economic effect should be described. The ratified rule can affect expectations immediately because buyers and lenders price future legal options. But completed-transfer effects require a live service. As of the cutoff, D3 was both an announced constraint and an unimplemented promise. It could chill decisions before it processed a single block, while the absence of transaction data prevented measurement of the size of that chill.

Why the bookkeeper test changes the answer

AFRINIC’s legitimate function is indispensable but narrow. Internet number resources must remain unique. Registry records should identify the recognised holder and contact. A transfer instruction should be authenticated. Fraud, duplicate claims and active disputes must be detected. WHOIS and RDAP information should be accurate. Reverse DNS and RPKI continuity should be protected. Court orders affecting a resource should be visible and obeyed. These are not ceremonial tasks; they reduce conflict and make independent networks safer.

None of them requires a geographic embargo. BGP does not validate a route according to whether a holder found a buyer inside AFRINIC’s service region. RPKI does not become cryptographically sound because an address originally came from one pool rather than another. A block remains unique after a cross-border sale if the old record is retired and the new control is correctly recorded. A service region tells an operator where registry administration is offered. It does not confer territorial title on the registry or on a vaguely defined “community”.

Heng Lu’s analysis calls this the moment when the bookkeeper mistakes the ledger for a throne. The distinction is not rhetorical decoration. A recordkeeper describes and authenticates a change in control. A sovereign can legislate, compel and punish under public law. AFRINIC is a private coordination institution. It is not a state, legislature, court, central bank or owner of the businesses whose number resources it records. It has no punitive jurisdiction over lawful commerce simply because its database is commercially important.

That does not mean AFRINIC must record any instruction placed before it. The registry can refuse an unauthenticated request, a duplicate claim, a transaction frozen by a competent court, a record with a known conflict or a transition that would break uniqueness. Those are objective failures tied to the integrity of the ledger. The registry can also coordinate timing with another RIR so both sides do not publish contradictory states. A narrow refusal preserves the purpose of registration.

Regional origin is different. It asks not whether the source controls the resource or whether the recipient record is accurate, but whether the resource was once drawn from AFRINIC’s pool. It then makes that history decisive for exit. The administrative label becomes an economic border even though AFRINIC acknowledges that it does not define a routing border. That is why “capital control” is the right institutional description. It is an analogy to the restriction of asset movement, not a claim that AFRINIC is a central bank.

The “community” cannot cure the gap by asserting collective ownership. Mailing-list participants and meeting attendees may develop useful standards. They do not thereby become the owners of member balance sheets or representatives of absent operators, investors, users and states. Consensus is evidence that a process reached its internal decision point. It is not a title deed over all African-administered IPv4.

The agency problem sharpens the concern. The authors, co-chairs, committee members, staff and directors who shape restrictions do not necessarily bear the same loss as a holder that cannot reach a buyer or use an address block as collateral. Their institutional incentives favour process, jurisdiction and continuity. A holder, lender or network customer carries the economic and operational downside. Good governance requires the decision surface to sit with those exposed to the consequences, except where a narrow common technical invariant truly demands collective action.

LARUS’s operational analysis treats transferability as a real property of IPv4 use, alongside routing visibility and registry reliability. Non-transferability can reduce flexibility and encourage leases, nominees or service arrangements that move economic control without a clean change in the registry. BTW’s research adds the market-architecture point: a one-way or uncertain path can deter inbound supply and create a regional policy discount. Neither source supplies a measured D3 price series. Their contribution is to establish the mechanism and the questions that evidence must answer.

The best case for keeping resources regional

The strongest good-faith defence of D3 begins with facts worth taking seriously. Africa received comparatively little IPv4 space. Networks still need IPv4 while deploying IPv6, and the pool is finite. Unrestricted global demand could bid away resources that a later-developing African network needs. Origin categories make complicated cases legible. Written approval can prevent double transfers and fraud. Reserved critical-infrastructure space may deserve special protection. A needs check can discourage speculative acquisition.

An inter-RIR system also requires reciprocal policies and staff coordination; it cannot function on seller preference alone.

D3, on this account, is a proportionate stewardship tool. It offers an inbound path, protects resources issued for the region, retains a path for Legacy and imported Global resources, and creates rules for accurate settlement rather than tolerating underground transfers. It spent years in public discussion, reached co-chair consensus, passed through the available appeal process, and was ratified after Board elections. ARIN later found it compatible. Because the Regional label does not change routing, the policy’s defenders can argue that no holder has been deprived of operational use.

This steelman gets several things right. Scarcity is real. Registry accuracy is valuable. A transfer that leaves two registries claiming different holders is dangerous. Proof of control, dispute checks and security continuity are necessary. A narrowly defined Reserved resource may need special continuity rules. Open discussion is better than an unpublished staff edict. Counterpart compatibility must be established rather than assumed.

But those premises do not justify the conclusion. Preserving an unallocated free pool for later applicants is an allocation rule. Restricting the exit of a resource already held and used by an operator is a rule over alienability. The first determines who receives common inventory; the second limits the market available to an existing holder. Calling both “stewardship” hides the change in function.

A regional wall does not create a single new IPv4 address or guarantee that a retained block will be deployed by the African network that most needs it. It can lower the price available to an African holder, reduce the value a lender recognises, and weaken the incentive to uncover unused capacity. It can also make an outside seller hesitate before moving a block into AFRINIC, even though D3 gives an imported Global block a later outward route on paper. That seller must still price implementation risk, possible policy change and corporate authority uncertainty.

The same legitimate safeguards can be achieved through less restrictive means. Verify the source. Authenticate the transaction. Reject duplicates and fraud. Publish the conflict status. Protect RPKI and reverse DNS. Coordinate simultaneous registry updates. Define a narrow rule for special-purpose resources. Provide reasons, deadlines and review for rejection. None of these functions requires Regional origin to decide whether an otherwise valid transfer may cross a service boundary.

Procedure also cannot carry more weight than it possesses. The Appeal Committee did not endorse D3 on its merits. The Board’s authority was contested, and the receiver’s discharge was unresolved. ARIN’s compatibility judgment did not implement AFRINIC’s system. Even a future court decision validating the corporate ratification would settle only a legal-authority issue within the case before it. It would not make geographic retention technically necessary or economically neutral.

The controlling answer is therefore disciplined rather than absolutist. AFRINIC should perform every check necessary to keep a truthful, unique and secure ledger. It may not treat the usefulness of that ledger as sovereign jurisdiction to confine capital. Regional scarcity supports transparent assistance, finance, leasing, market access and—where an actual public authority chooses—public subsidy. It does not supply a private registry with title over holders’ exit choices.

Economic direction without invented precision

The basic effect follows from options. A Regional holder can sell through the intra-AFRINIC path but cannot, under D3, use the ordinary outward inter-RIR path. A Global or Legacy holder can reach a wider set of eligible counterparties. All else equal, a broader option set cannot be worth less than a narrower one. In practice, block quality, reputation, size, legal risk, timing and buyer concentration may dominate a particular deal, so theory alone cannot state the size—or even the observed sign in every local comparison—of a price difference.

Liquidity is the first channel. Excluding buyers can reduce bid depth, widen spreads and increase the time required to clear a transaction. Financing is the second. A lender may haircut a block when registry-recognised transfer depends on geography, staff discretion, unresolved implementation and litigation. Inbound supply is the third. Capital dislikes uncertain exit and may demand a discount before entering. Transaction cost is the fourth: both registries must coordinate, status and control must be proved, and legal review can lengthen the path.

Informal substitution is the fifth and most corrosive channel. If parties cannot record a sale, they may lease addresses, use nominees, delegate routing or wrap control in service contracts. Some of those arrangements are lawful and useful. The danger appears when the economic controller, the registry contact, the route origin and the security authorisation drift apart because a clean registry update is too risky or unavailable. WHOIS and RDAP then describe yesterday’s relationship; incident responders contact the wrong party; RPKI and reverse DNS require awkward workarounds.

A restrictive registry can defeat registry accuracy in the name of protecting it.

No transaction dataset in the reviewed evidence permits an exact D3 impairment estimate. Heng Lu’s February analysis used values of about USD 45,000 for a /22 and USD 2.88 million for a /16. Those numbers are arithmetically consistent with a common assumed price of USD 43.9453125 per address: a /22 has 1,024 addresses, while a /16 has 65,536, exactly 64 times as many. They are block-value scenarios, not measured losses caused by the policy.

The honest calculation is conditional. If a block contains N addresses, the assumed unit price is p, and the mobility discount is d, an estimated impairment is N × p × d. Under the price implied above, a /22 scenario would be USD 45,000 × d and a /16 scenario USD 2.88 million × d. No observed value for d is available. Setting d to one would assume the entire block value disappeared, an extreme proposition for which the record provides no support.

The same caution applies to the proposal’s statement that the AFRINIC pool represented 7.23 /8s. Multiplying 7.23 by 16,777,216 yields approximately 121,299,272 addresses. That is a conversion of fractional-/8 accounting, not an audited current inventory. Without a category breakdown, it cannot tell us how many addresses are Regional, how many holders want to leave, or how much capital is affected.

The evidence therefore closes direction more firmly than magnitude. D3 narrows the formal exit option of Regional resources. A narrower option can affect liquidity, collateral and bargaining power. But the actual discount, volume, spread, processing time and failed-deal count remain unknown. Serious scrutiny should demand bids, comparable transactions, time-to-close, rejection reasons, collateral haircuts and transfer logs—not replace missing data with confident decimal places.

What can be concluded now

Four findings are firm. First, the text creates an origin-based asymmetry: Regional stays inside the ordinary inter-member market, while Global and Legacy have stated paths out. Second, AFRINIC ratified that text on 4 February 2026 and publicly confirmed that AFRINIC-issued IPv4 could not leave through the policy. Third, the policy’s internal procedural challenge ended on admissibility rather than merits, while its corporate authority was later placed directly before a court. Fourth, ratification had not produced a publicly demonstrated operational inter-RIR path by 10 August.

Several conclusions are not available. The public record does not show the Board’s exact resolution, vote or receiver relationship. It does not disclose the full Skyconnect pleadings or a final merits judgment. It does not establish the receiver’s final status after the latest update. It does not close APNIC and LACNIC activation, practical ASN treatment, category inventories or market demand. It does not prove any particular loss, discount or volume of workaround arrangements.

The institutional judgment does not depend on filling those gaps with speculation. AFRINIC is a ledger keeper and coordinator, not a sovereign. Its service region is an administrative service area, not a border around title. Accurate transfer recording is legitimate; a geography-based veto over an otherwise valid transaction is capital lock-in. The rule can be exact, publicly discussed and corporately ratified while still exceeding the narrow function that makes registry cooperation legitimate.