Summary

  • The practical IPv4 market is not global merely because addresses are globally routable. A cross-region transfer exists only where the source and destination registries recognise the resource type, accept each other's conditions and coordinate the record change. The result is a changing matrix of permitted routes rather than one market.
  • Compatibility can protect legitimate interests: confirming the registered holder, checking that a block is dispute-free, preventing duplicate registration, authenticating the recipient and synchronising registry, RPKI and reverse-DNS changes. Those tasks are tied to accurate custody. A second registry's demand forecast, regional development policy or institutional preference is not automatically part of that minimum.
  • ARIN currently requires a reciprocal, compatible, needs-based counterpart and applies a 24-month recipient need test. APNIC assigns source conditions to the source registry and recipient conditions to the recipient registry. RIPE NCC supplies a five-year, 50% use plan when another registry demands a needs-based counterpart. LACNIC limits transfers to IPv4 and applies its own holding and eligibility periods. These differences create pair-specific costs and vetoes.
  • AFRINIC's February 2026 announcement says an inter-RIR policy has been ratified, with reciprocal transfers for eligible categories, while RIPE NCC's current operational page still says that AFRINIC has no usable inter-RIR policy. The discrepancy is not a reason to choose one institution's description. It shows that policy approval, counterpart recognition and an executable transfer route are different facts.
  • A minimum standard should use home-rule allocation: the source registry verifies the source and resource; the destination registry verifies the recipient and resulting record. A registry should not export unrelated domestic policy as a condition of recognition. Common transfer packets, published compatibility decisions, reason codes, time limits, appeals and coordinated technical cutover would preserve integrity without market segmentation.
  • Number Resource Society offers a positive direction if it remains a portable ledger service and compatibility monitor rather than another permission centre. It can publish the pair matrix, test record portability, compare delay and rejection data, preserve signed evidence and advocate a narrow recognition standard while leaving price and commercial choice to address holders.

A trade can clear commercially and fail administratively

Consider a network in one region that has agreed to buy a clean IPv4 block from a registered holder in another. The buyer has budget approval, the seller has corporate authority, a broker has checked sanctions and title history, and an escrow arrangement protects payment. The buyer's transit providers are prepared to announce the prefix. The contract makes completion conditional on the authoritative registration moving to the buyer.

There is still no completed transfer.

The source Regional Internet Registry must accept that the seller may release the block. The destination registry must accept that the buyer may receive it. Each may ask the other to certify facts. They must agree on the resource class, any holding period, the meaning of demonstrated need, the recipient's regional connection, the treatment of legacy status and the date on which their records change. If one institution says the other's policy is not reciprocal or compatible, the contract cannot produce the record on which the parties depend.

This is an unusual border. An IPv4 address has no region bit. BGP does not inspect the buyer's registry membership before carrying a route. A prefix recorded in one region can be used in another, subject to network acceptance and applicable law. Geography enters through administrative custody: which registry holds the authoritative record, which account can request changes, and which institution is willing to recognise a transfer from its counterpart.

That administrative boundary can be useful. Two registries changing the same prefix independently would destroy confidence. A forged seller should not be able to move space. Existing Route Origin Authorisations, reverse-DNS delegations and routing records require an orderly cutover. Commercial agreement alone cannot answer those questions.

But record protection does not explain every restriction. A destination may require a forecast of address use. A source may impose a waiting period because it previously allocated scarce space. A registry may refuse an entire region because its counterpart lacks a policy carrying the right words. The same buyer and seller can therefore complete one regional route but not another even when the evidence of identity and custody is identical.

The institutional question is not whether registries should inspect anything. It is which inspections are necessary for a truthful, unique and recoverable ledger, who applies them, and whether a difference unrelated to those ends may block the trade. Compatibility becomes legitimate when it is a bridge between records. It becomes a private trade barrier when it turns domestic policy variation into an external veto.

This is a private barrier, not a claim about customs law

Calling incompatibility a trade barrier needs precision. Regional Internet Registries are not customs authorities, their policies are not tariffs, and an address transfer is not an ordinary shipment of goods. The point is functional: private institutions control a registration change without which many buyers will not close, and their mutual-recognition rules determine which counterparties can transact across regions.

The barrier has three features. First, it is categorical. A willing buyer cannot pay for additional inspection when no recognised path exists. Second, it is pair-specific. The same resource may be transferable between two regions but not between another pair. Third, it is embedded in recognition. Parties can sign around many commercial terms, but they cannot privately create an authoritative recipient record in a registry that refuses the transfer.

That combination resembles a non-tariff barrier more than a service fee. Its cost appears as abandoned negotiations, narrower bidder pools, legal conditions that never become effective, duplicate due diligence, longer escrow, broker uncertainty and the discount attached to a block with fewer possible destinations. None of those costs appears in a successful-transfer log. The blocked route leaves no public row.

The phrase also identifies who bears the burden. A large platform can maintain entities in several regions, pre-qualify at more than one registry and divide a purchase among experienced advisers. A smaller operator may have only one recognised organisation, one intended network and a fixed deployment date. If the direct pair is closed, corporate restructuring or a staged transfer may be too expensive, unavailable or inconsistent with policy. The nominal rule applies equally while the incidence does not.

Nor is the barrier made public simply because policy discussion occurred on an open mailing list. The decisive act is performed by incorporated associations and their staff through account, contract and registry control. An operator that did not participate still depends on the outcome. Openness of debate can improve reasons; it does not convert regional policy into a universal public mandate.

This distinction prevents overclaiming. The argument does not require treating number resources as unrestricted property, and it does not say every proposed sale must be recorded. A registry can reject forgery, disputed custody, duplicate claims and an unsafe cutover. It can apply the recipient obligations that govern records in its own service region. What it should not do is use the word compatibility as a blank cheque for one private institution to enforce a broader economic policy outside its own record.

The test is therefore narrower than market liberalisation. Ask whether the condition proves who may change the record, keeps one authoritative state, preserves operational continuity or creates a reviewable recipient relationship. If not, the institution must identify the separate authority that allows it to close a cross-region route.

There is a pair matrix, not a single world market

The cleanest description of inter-RIR trade is a directed graph. Each registry is a node. An edge exists only for a specified resource class, direction and date, and only after both endpoints accept the path. "Inter-RIR transfers are allowed" is too coarse to describe that graph.

RIPE NCC's current inter-RIR transfer page illustrates the point. It lists IPv4 and AS Number transfers, including legacy resources, between RIPE NCC and ARIN, and between RIPE NCC and APNIC. It lists IPv4, including legacy resources, between RIPE NCC and LACNIC. It says AFRINIC is not currently available for transfers. ARIN's transfer guidance likewise lists APNIC, LACNIC and RIPE NCC as approved counterparts, excludes AFRINIC, permits IPv4 and AS Numbers with some counterparts, and excludes IPv6 from ARIN inter-RIR transfers.

The edge is therefore typed. RIPE NCC policy permits transfers of IPv4, IPv6 and AS Numbers inside its region, but that does not create an inter-RIR IPv6 path through ARIN. LACNIC's transfer rule concerns IPv4, so its compatibility with a registry that can move AS Numbers does not make AS Numbers transferable on that pair. Legacy status can survive one destination or be converted under another contractual choice. A statement about the pair is incomplete without the resource and resulting status.

Direction matters too. Source conditions and recipient conditions are not the same. A block may be eligible to leave because its holder has passed a source waiting period, while the buyer fails the destination's need or regional-presence rule. In the reverse direction, a different block and buyer face a different combination. A table with an undirected check mark conceals that asymmetry.

Time is the fourth dimension. Policies change; staff documents follow; counterparts assess compatibility; systems are tested; forms and secure contacts are established. An edge can exist in policy language before the registries are ready to use it, or remain described as closed on one registry's page after another announces ratification. Commercial parties need the executable state on the intended closing date, not a historical vote.

A defensible public matrix would thus have one row per directed pair and resource class. It would show source rule version, destination rule version, status, effective date, allowed holder classes, minimum block, waiting periods, need test, legacy treatment, required contracts, technical cutover obligations, median completion time, appeal route and last successful test. Every compatibility decision would link to reasons and a responsible contact.

No RIR should own the interpretation of the whole graph. Each can state its own edge, but the two statements must reconcile. A neutral publication can expose disagreement without pretending it has transferred authority. The first transparency reform is simply to stop presenting a changing matrix as though it were a natural property of the Internet.

Four policy designs produce different gates

The present routes are not variations in wording around one common rule. They assign different powers to source and destination registries.

ARIN's model is the most explicit mutual-recognition gate. Its guidance says an inter-RIR transfer requires reciprocal, compatible, needs-based policy. An ARIN recipient must demonstrate up to a 24-month supply and, for a request above an initial /24, document that 50% will be used within 24 months along with utilisation of prior holdings. An ARIN source must be the registered holder, be free of dispute, meet a /24 minimum, observe a 12-month restriction after certain receipts, and accept a 36-month waitlist consequence after transferring IPv4 space. ARIN may seek certification from the receiving registry that its review is compatible.

APNIC's active Internet Number Resource Policies take a home-rule approach in form. Inter-RIR IPv4 transfers require a counterpart policy that permits the exchange. Conditions on an APNIC source are set by APNIC; conditions on an external source come from its registry. Conditions on an APNIC recipient are set by APNIC; conditions on an external recipient come from its registry. APNIC recipients still face a demonstrated-use plan, and addresses delegated from the 103/8 pool are subject to a five-year transfer restriction. The architecture is separable even where the substantive tests remain interventionist.

RIPE NCC acts partly as a compatibility adapter. RIPE-807 allows a wide set of resources to move and generally follows its internal transfer policy, but adds a special bridge where the source region demands needs-based reciprocity: a recipient in the RIPE NCC region provides a plan to use at least 50% of the resource within five years. The current operational page also requires at least one active network element in the RIPE NCC service region for incoming resources and usually a contractual relationship, with an exception for legacy resources.

LACNIC's policy manual permits intra- and inter-RIR IPv4 transfers. A LACNIC recipient must justify the address space; an external recipient follows its own registry's criteria. The holder and dispute status are verified. The source becomes ineligible for new IPv4 resources from LACNIC for one year, a transferred block cannot move again for one year, and addresses originally allocated or assigned by LACNIC cannot be transferred for three years. Incoming legacy resources lose legacy status.

These are policy choices, not clerical translations. A 24-month utilisation forecast and a five-year plan are not the same evidence. A three-year origin lock and a five-year 103/8 lock change which supply can reach the border. Preserving legacy status and extinguishing it change the recipient's contractual position. "Compatible" can therefore mean exact reciprocity, an accepted substitute, a source-destination division of labour or a political judgement that differences are tolerable. Without a published minimum, the label explains the result only after the registries have chosen it.

AFRINIC shows why ratification is not execution

The African route makes the temporal problem impossible to ignore. AFRINIC announced that its Number Resources Transfer Policy, AFPUB-2020-GEN-006-DRAFT03, was ratified on 4 February 2026. The announcement says the policy covers intra-regional, legacy-to-member and reciprocal inter-RIR transfers of IPv4 and AS Numbers. It also says AFRINIC-issued IPv4 cannot leave the region, while some categories, including certain legacy and externally received resources, may be eligible for outbound movement.

That is a material policy event. It is not proof that every counterpart has opened an operational edge. RIPE NCC's current transfer page, retrieved months later, still says AFRINIC has no inter-RIR policy and that resources cannot move to or from the region. ARIN's current page also lists AFRINIC as not approved. The public record therefore contains at least three distinguishable states: AFRINIC has announced ratification; the announced policy is selective; and major counterpart pages do not yet describe an approved route.

The correct response is not to declare AFRINIC open because its announcement is newer, nor to declare the ratification unreal because counterpart pages lag. A buyer needs evidence of implementation, reciprocal recognition and a tested exchange for the specific resource. Until that evidence exists, the route is commercially uncertain.

This gap is itself a barrier. Parties cannot price closing risk from a policy headline. Does the source category qualify? Has the counterpart completed legal review? Which institution begins the request? Are secure contacts and data-sharing terms in place? What happens to RPKI state? Has a first case completed? Is there a service target? If the registries disagree, where can either party obtain a reason and review?

The asymmetry also exposes the danger of regional conservation rules. Preventing newly issued AFRINIC space from leaving may protect a remaining pool from immediate extraction. Applying a permanent regional export restraint to every address, including independently acquired or legacy space, would be a different claim. AFRINIC's own summary draws categories, which should be preserved in analysis rather than compressed into "open" or "closed."

A minimum compatibility standard must therefore define readiness. Ratification is the policy-authority milestone. Implementation requires published procedure, counterpart acceptance, technical exchange, staff readiness and a start date. Operational availability requires a request channel and service reporting. Proven availability requires a completed or independently tested case. Each status should be public.

The lesson extends beyond AFRINIC. A market cannot depend on two institutions silently updating web pages after bilateral discussions. Compatibility is a public fact with private economic consequences. It needs a versioned record, an effective date and reasons strong enough for a seller to put at risk a valuable block and for a buyer to place funds in escrow.

Mismatched needs tests export domestic policy

Needs assessment is the central compatibility fault because it can travel across the border. A destination registry may reasonably ask which organisation will hold the record and whether it accepts the destination's ongoing terms. Requiring a forecast of commercial use is more controversial. When a source registry refuses to recognise a destination unless that destination performs such a test, a regional rationing policy becomes a condition on an external buyer.

ARIN's 24-month standard is concrete enough to see the effect. A buyer may have negotiated a block because it expects uncertain customer growth, wants resilience inventory, plans acquisitions, or needs capacity beyond a two-year horizon. The transaction price already gives the buyer a reason not to acquire useless space. ARIN can still believe market price is limited public evidence protection against speculation. But that belief does not establish why a RIPE NCC, APNIC or LACNIC buyer must satisfy ARIN's allocation philosophy before an ARIN source can update its record.

RIPE NCC's five-year, 50% plan was designed to make the route compatible without importing the exact 24-month test. It is a pragmatic bridge, but it reveals the institutional trade. A destination that otherwise accepts transfers has created a special review because another region conditions recognition on needs-based policy. The buyer's burden comes from the source border, not solely from the rules ordinarily attached to the destination record.

This can segment demand. Buyers with predictable, document-heavy deployments can qualify across more routes. Buyers with volatile growth, new business models or limited administrative capacity face a narrower market. Brokers screen not only ownership and routing reputation but the probability that two registries will accept the demand narrative. A rule intended to deter hoarding can favour incumbents able to produce conventional forecasts.

The record-integrity alternative is narrower. Verify the legal recipient, account authority, absence of conflicting claims, size and identity of the block, and acceptance of destination obligations. Publish the transfer. Apply ordinary anti-fraud and competition law where relevant. If a destination community wants a needs test for records under its custody, it can impose and defend that rule on its recipients. The source registry should not demand equivalence unless it can show that the destination's omission would corrupt the reference or create duplicate use.

This does not make every acquired address permanently unrestricted. Holding periods tied to a recent subsidised or scarcity allocation can prevent immediate arbitrage of that allocation. Contractual obligations can survive where clearly attached. Sanctions, court orders and disputes can block a case. The line is purpose and incidence: conditions should follow the institution's own resource grant, its own holder or the integrity of the shared cutover, rather than become a universal economic licence.

Compatibility should mean that two different systems can exchange a trustworthy state, not that one system must imitate the other's theory of demand.

Resource class and legal status can close an otherwise open route

Even if needs tests were harmonised tomorrow, the market would remain segmented by what may move and what the resource becomes.

IPv4 is the principal traded resource because exhaustion created scarcity and because all currently active pairs address it. AS Numbers can move on some paths, including between RIPE NCC and ARIN or APNIC, but not automatically through an IPv4-only counterpart. IPv6 transfer rules exist within some regions, yet ARIN expressly excludes IPv6 from inter-RIR transfers. A broker's statement that two registries are compatible is meaningless without naming the resource class.

Status is equally important. RIPE NCC permits an incoming legacy resource to retain legacy status. A recipient can choose a contractual relationship for service and RPKI eligibility, but the transfer page says such a relationship is not required merely to receive a legacy block. LACNIC's manual says legacy resources transferred into its region cease to be legacy. The same block can therefore carry a different post-transfer relationship depending on destination.

That consequence can change value. A buyer may care about annual fees, contractual warranties, dispute provisions, certification access, sub-assignment rights and future transferability. A seller may warrant only that the present registration is accurate, not that a destination will preserve status. If the status change is discovered after price agreement, the deal can be renegotiated or abandoned.

Holding periods create provenance-dependent supply. APNIC's 103/8 restriction follows the allocation date. LACNIC imposes a three-year restriction on its allocations and a one-year restriction on retransferring a received block. ARIN prevents certain sources from transferring within 12 months after receiving space. RIPE NCC restricts scarce resources for 24 months after receipt, subject to defined exceptions. A block's route is therefore not determined by current holder and destination alone; its acquisition history matters.

A reliable compatibility service must carry those attributes as evidence, not force parties to infer them from scattered records. The source registry should issue a signed eligibility statement identifying the resource, holder, status, relevant origin date, active locks, disputes and permitted destination classes. The destination should issue a signed acceptance statement identifying resulting status, contract, services and restrictions. Neither statement should disclose price.

This packet would reduce repeated interpretation without erasing policy. More importantly, it would show which restrictions are genuinely attached to the resource and which are merely the counterpart's general preference. Parties could challenge an incorrect classification before money and routing plans reach the closing date.

Portability begins with knowing what is portable. A global ledger standard need not decree that every status survives. It must make any transformation explicit, justified and visible before transfer.

Coordination protects the record but can also conceal delay

An inter-RIR transfer is operationally harder than an internal one. RIPE NCC says both registries must approve the request and synchronise their updates. Its guidance warns that coordination across time zones and compliance with both policies take longer. ARIN notes that duration can vary because several registries and organisations are involved. Those statements are credible descriptions of extra work. They are not measurements of whether the extra time is necessary.

The cutover has real dependencies. The source registry must remove or change the holder. The destination must create the recipient state. Existing RPKI certificates and Route Origin Authorisations may be revoked or shrunk; the recipient then needs new certification where eligible. Reverse-DNS delegation and Internet Routing Registry entities may require action. Public RDAP and Whois responses must stop pointing to the former holder. The parties need a date that does not leave simultaneous authority or an unexplained gap.

ARIN's source checklist tells holders to update ROAs, IRR objects and reverse-DNS plans. RIPE NCC explains that outgoing resource certificates are revoked or reduced and that new certification depends on the destination. These are record-continuity duties, precisely the domain in which compatibility is justified.

Yet a two-registry process can make delay ownerless. Each party sees only its ticket. ARIN says source and recipient requests are processed independently and protects each organisation's confidentiality, so the parties should coordinate directly. A pause can arise from the seller's documents, the buyer's need evidence, the source analyst, the destination analyst, the handoff, fees, contract signature or a disputed technical entity. A completion date reveals none of that.

The minimum standard should define a shared clock without publishing private documents. Each case receives a common identifier known to both registries and parties. The public aggregate records timestamps for accepted request, source verification complete, destination verification complete, handoff sent, cutover scheduled and cutover complete. Time waiting for a party is separated from registry-controlled time. Pauses have standard reason categories. Percentiles are published by directed pair and resource class.

A failed synchronisation should have a rehearsed recovery plan. The registries should identify the authoritative pre-cutover state, preserve signed snapshots, specify who can stop the change, and define how RPKI and reverse DNS are restored. This is not an argument for casual rollback after commercial completion. It is protection against a partial administrative change.

When coordination is observable, extra time can be defended. When it is not, complexity becomes a self-validating explanation and the parties pay for a delay no institution has to own.

Measure the barrier by paths that never complete

Successful transfers are the visible tip of compatibility. They prove that an edge worked for a qualifying case. They do not show how much trade the rule displaced.

A serious assessment begins with attempted paths. For each directed registry pair and resource class, count expressions of interest that became complete requests, requests accepted for review, approvals, rejections, withdrawals and expiries. Record the amount of address space at each stage without publishing applicant identity where the case did not complete. Distinguish policy ineligibility, missing holder evidence, dispute, need shortfall, holding period, regional-presence failure, counterpart incompatibility, sanctions, party withdrawal and commercial failure.

The denominator matters. If a pair reports 100 completed transfers and no failed requests, observers cannot tell whether approval is predictable or whether brokers filtered 900 plausible deals before filing. Registry data alone will not capture every negotiation, but pre-qualification and listing services can report anonymised attrition. Brokers can submit audited aggregates without naming clients or disclosing contracts.

Delay should be measured by stage and incidence. Median alone is limited public evidence because a market deadline is often lost in the tail. Publish the 25th, 50th, 75th and 90th percentiles, pending-case age bands, and the share exceeding the service target. Split party-controlled days from registry-controlled days and bilateral handoff days. A pair whose median is quick but whose oldest cases disappear for months has a material barrier.

Price effects can be studied without requiring transaction-level publication. Compare bids or realised prices for otherwise similar blocks by eligible destination set, status, reputation, size and expected closing time. A block transferable to four regions should not be assumed equivalent to one confined to a single region. The difference is not wholly caused by policy, but the route set is an observable variable.

The most direct indicator is the compatibility coverage ratio: eligible address supply with an executable route to a stated destination divided by potentially transferable supply under source custody. It should be computed for each resource status and age. A second indicator is buyer coverage: verified recipient organisations able to reach each source region under current tests. These are not measures of whether every trade is desirable. They measure whether geography, rather than record risk, decides access.

Every estimate needs caveats. Private negotiations are incomplete, broker samples are selective, failed parties may have weak documents, and address reputation affects value. The answer is not to publish a false global rejection rate. It is to expose the observed stages, methods and missing population so that institutional claims can be tested rather than repeated.

The minimum standard should be narrow enough to travel

A global compatibility constitution will fail if it tries to merge every regional policy. The workable standard is a small set of duties that protect the shared record while allowing domestic rules to remain domestic.

First, recognise the legitimate source and recipient. The source registry authenticates the current holder, corporate authority and absence of a recorded dispute. The destination authenticates the recipient and establishes the relationship required for the resulting record. Each accepts the other's signed finding unless it identifies a specific evidentiary defect.

Second, preserve uniqueness. Both registries agree on the exact resource set, any excluded sub-block, status, active lock and transfer date. A cryptographic digest of the agreed packet prevents silent alteration. The destination cannot activate before the source commits to release, and the source cannot close without destination acceptance.

Third, keep home rules at home. Source restrictions tied to the source's own grant, holder or court obligations may apply. Destination restrictions tied to the destination's recipient and ongoing services may apply. A source registry does not require the destination to adopt a general needs regime merely to prove record accuracy. A destination does not re-judge the source's corporate title without identified contrary evidence.

Fourth, make status transformation explicit. Before approval, the recipient receives a statement of resulting legacy or contractual status, fees, certification eligibility, retransfers, sub-assignment rights and required services. No material status change appears for the first time at cutover.

Fifth, coordinate operational state. The packet includes RPKI, RDAP, Whois, reverse-DNS and relevant routing-record actions, responsible parties, target times and the safe pre-cutover state. Registries publish completion service levels and test their exchange periodically.

Sixth, give reasons and review. A refusal identifies the clause, facts found, missing evidence and correction route. A party can seek review by a person not responsible for the first decision. Bilateral disagreement has a defined escalation with a time limit; neither registry can leave the case indefinitely in the other's queue.

Seventh, publish comparable aggregates. Directed-pair volumes, failures, reasons, stage times, pending age and technical incidents are released under a common definition. Commercial price and personal documents remain private, but the public can see whether compatibility performs.

The standard should be testable. A synthetic transfer packet using reserved examples can be exchanged quarterly. Results show whether fields, signatures, contacts and cutover messages still interoperate. A failed test does not move a real resource, but it prevents the first live case after a policy change from becoming the test.

This design is intentionally less ambitious than uniform policy. Its legitimacy comes from restraint. It asks each registry to do what a reliable ledger requires and to justify anything more in the jurisdiction where that power is exercised.

Fraud, scarcity and autonomy do not justify an unlimited veto

The strongest objection is fraud. IPv4 blocks are valuable, corporate histories are complex, contacts can be stale and an incorrect transfer can disrupt a network. If mutual recognition narrows review, a forged source might exploit the weakest registry.

The answer is a common evidence floor, not policy identity. Corporate existence, signatory authority, historic registration, disputes, sanctions, court orders and chain of control are verifiable subjects. Registries can require stronger evidence when a specific risk is present and share it under lawful confidentiality. They can publish audit results and reversal incidents. A demand forecast does not prove that the seller owns the block; a regional lock does not authenticate a director.

A second objection is scarcity policy. Regions that received later allocations may fear that a global market will export address space to wealthier buyers. That distributional concern is real, especially where historic allocations were unequal. But a geographic export ban should be defended as a redistribution rule, not hidden inside technical compatibility. Its scope, duration, affected resource classes, evidence of effect and review date should be explicit. Newly allocated scarcity space can carry a time-limited condition without treating every legacy or purchased block as regional capital forever.

A third objection is community autonomy. Each RIR developed policy through its own process, and a minimum global rule could displace local participation. Yet autonomy cannot mean freedom to impose domestic conditions on every external holder without explanation. Home-rule allocation preserves more regional choice than exact reciprocity: each registry governs its recipient and its own grants, while the shared standard governs only the handoff.

A fourth objection is forum shopping. Buyers may choose a destination with lighter conditions and use the addresses elsewhere. Current practice already recognises that registry region and network use can diverge, while some registries require a regional connection. The response should be a clear, auditable destination nexus and truthful registration, not an attempt by the source to regulate all future routing. If a destination's nexus rule is too weak, counterpart registries should identify the concrete record or legal risk.

Finally, critics may say registries are not market regulators and should not facilitate trade. The minimum standard does not ask them to set prices, broker parties or guarantee contracts. It asks them not to make accurate records conditional on unrelated institutional preferences. Refusing to record a real transfer does not make the underlying operational or economic change disappear; it can make the public record less true.

Restraint works in both directions. Registries should not promise that a clean record proves property title, route acceptance or commercial performance. Traders should not demand registration of a disputed or unsafe change. A narrow standard protects that boundary better than a compatibility label broad enough to mean anything.

NRS can make portability practical without becoming another gatekeeper

Number Resource Society is most useful here as a direction for institutional design, not as a sixth regional authority. The gap is not a shortage of organisations able to say yes or no. It is the absence of a neutral, operator-centred record of which paths work, why they fail and whether a holder can carry verified state between custodians.

NRS could maintain the public compatibility matrix as evidence. Each row would cite the current policy, implementation notice, counterpart confirmation, last test, permitted resource classes, status effects and performance aggregates. Conflicting institutional claims would appear side by side until resolved. NRS would not announce that a path is authoritative merely because one registry says so; it would state the evidence level.

It could also define and test a portable transfer packet. The packet would include holder identity assertions, resource set, chain-of-custody references, dispute state, applicable locks, requested destination, resulting-status statement and technical cutover checklist. Sensitive documents would remain encrypted for the parties and registries. Public proofs would show issuer, date, integrity and revocation without exposing passports, contracts or price.

Portability is not a promise that any holder can bypass current authority. It is the practical ability to export a complete, authenticated record so the next recognised custodian does not force the operator to reconstruct history from email and screenshots. The less dependent the evidence is on one institution's private case file, the lower the exit cost and the weaker the opportunity for delay to become leverage.

NRS could publish comparative transparency: completion and failure rates, registry-controlled time, handoff time, reason categories, appeal outcomes, data freshness and cutover incidents. It could invite operators and brokers to report anonymised pre-filing attrition, with methods and sample limits. This would reveal the shadow cast by closed paths without claiming access to every private negotiation.

The non-gatekeeping limit is essential. NRS should not charge for a certificate that registries must accept, decide whether a buyer needs addresses, choose winners among brokers or speak for holders without a revocable mandate. Its value comes from reducing information asymmetry, preserving evidence and advocating mutual recognition. If its approval becomes another mandatory step, it has reproduced the barrier.

This role also disciplines the positive case for NRS. A portable ledger service must publish export formats, service targets, correction rights, succession arrangements and independent audits. An operator should be able to leave NRS without losing its evidence. The institution proves its alternative by making itself less indispensable.

The direction is therefore not market triumph over coordination. It is coordinated records without inherited territorial control: a thin service that keeps one truthful state, lets verified evidence travel and refuses to monetise permission.

Compatibility is legitimate only when refusal is explainable

Inter-RIR transfer policy sits at the boundary between technical coordination and economic power. The registry's technical task is real: preserve unique records through a risky change. The economic effect is also real: a refusal can remove an entire region of buyers or sellers. Institutional legitimacy depends on acknowledging both.

The current system often treats compatibility as a binary conclusion delivered by registries after comparing policies. That is inadequate for a market in which the conclusion changes asset reach. The comparison must identify the necessary record controls, the differences that remain, who bears each condition and why a difference is serious enough to close the route.

A published reason can be challenged. "The destination cannot authenticate the recipient" identifies an evidence problem. "The resource class has no receiving record" identifies a technical or policy gap. "The counterpart lacks a needs-based policy" identifies a philosophical difference, but it does not yet explain why the reference would be false. Precision turns compatibility from institutional deference into a reviewable decision.

The same standard should govern openings. A press release that celebrates global alignment is not enough. Parties need an effective date, forms, service targets, status consequences and a tested cutover. The AFRINIC record in 2026 makes this especially clear: ratification is important, counterpart recognition is separate, and executable service must be demonstrated.

The market should not demand complete uniformity. Regions can keep different membership systems, fees, internal transfer rules and recipient duties. What they cannot plausibly claim is that global uniqueness requires every regional preference to travel with a prefix. Uniqueness requires one current record and a reliable transition. It does not require one private constitution for all address use.

The minimum standard therefore protects registries by narrowing their promise. They authenticate, record, publish, coordinate, correct and hand over. They do not guarantee price, commercial wisdom, route acceptance or the buyer's business plan. A registry that performs the narrower role well has a stronger claim to trust than one whose broad discretion cannot be separated from record custody.

For buyers and sellers, the reform turns a hidden condition into a known route. For operators, it reduces the risk that administrative geography interrupts deployment. For members, it creates evidence with which to judge policy. For registries, it replaces ad hoc bilateral interpretation with a defensible service boundary.

An IP address is globally usable because networks coordinate, not because private borders are natural. Inter-RIR compatibility should carry the record across those borders. It should not be the mechanism by which the borders are made permanent.

Sources