Summary
- Internet number resources required globally coherent state. The surviving evidence does not establish that coherence also required every holder to depend permanently on one regional registration provider.
- Regional exclusivity had a substantial case: identifiable responsibility, local language and custom, fair sub-allocation, administrative scale and compatibility with provider-based allocation and routing aggregation.
- Registry-service portability would have changed only the servicing institution. It could not legitimately change the holder, duplicate the resource, erase allocation conditions, confer provider-independent routing or conceal a transfer.
- The 1998 DNS policy comparison showed that coordinated uniqueness and plural registration service were institutionally separable in a related identifier system. It did not demonstrate that routed address resources could support the same arrangement safely.
- A workable handoff would have required authoritative finality, inherited obligations, fraud-resistant authentication and continuity across reverse DNS, failure recovery, later security functions and legal disputes.
- The historical material contains no comparable service dataset, production specification, interoperability trial, cost model or failure exercise for such an arrangement.
- The bounded finding is therefore two-sided: uniqueness does not by itself prove that permanent service attachment was indispensable, but conceptual separation does not prove that portability was feasible, safe or beneficial between 1992 and 2005.
The question that the 1992 plan did not settle
Internet address administration began from a constraint that was not negotiable. Globally significant number resources could not be validly assigned to incompatible claimants at the same time. The system needed a coherent account of which resources had been delegated, to whom, under which conditions and through which chain of authority. Without that coordination, address uniqueness would have failed at the point where separate administrative decisions collided.
That necessity did not answer every institutional question surrounding the record. In particular, it did not establish whether the organisation administering a holder’s entry had to remain the same for as long as the holder retained the resource.
The opening for this question appears in RFC 1366, published in October 1992. The proposal favoured geographically distributed administration and one registry for each region. It gave practical reasons for doing so: regional institutions could work with local language and custom, contribute to fair sub-allocation and reduce the burden on the central Internet Registry. Those considerations formed a serious case for regional service rather than a merely territorial preference.
The same document nevertheless preserved a route for a subscriber to approach the central IR directly when necessary. That route was not lateral portability. It did not grant a holder a right to choose among regional registries, turn the central IR into a competing service provider or explain how responsibility for an existing allocation would move. It supplied neither an ordinary switching procedure nor rules for ending the former registry’s authority.
Its narrower importance is that administrative contact was not described as absolutely inseparable from geography in every circumstance. The regional institution was the preferred service point, but the wider hierarchy could accommodate another administrative route. That limited observation is sufficient to ask whether the arrangement could have been structured differently without claiming that portability was already intended.
The counterfactual changes one relationship. It keeps a globally coherent record and asks whether a holder could have replaced the registry servicing that record while keeping the same resource, the same holder identity, the same allocation history and the same continuing obligations.
The question is not whether separate registries could maintain competing versions of allocation truth. They could not do so without undermining uniqueness. Nor is it whether an address block could follow its holder through every change of network provider or topology. That would raise a different routing question. The issue is whether the validity layer and the service layer could have been organisationally distinct.
The historical record cannot supply a direct comparison. There was no portable address-registry regime operating alongside the regional model during the period. The available sources describe proposals, operational guidelines, a related institutional comparison, recognition policy, scholarly analysis and a much later portability argument. They permit a controlled examination of the constraints, but not a finding about an alternative history that never occurred.
The resulting inquiry is therefore narrower than asking whether the Internet chose correctly in 1992. It asks what regional exclusivity addressed, which functions genuinely required common authority and what a replaceable service relationship would have needed to preserve.
Porting a service relationship is not moving an address in every sense
The word “portability” becomes misleading when it is allowed to cover several different changes.
Registry-service portability would mean that Registry B replaces Registry A as the recognised institution providing administrative service to an unchanged holder for an unchanged resource. The common record would show the new servicing relationship while retaining the allocation chain, applicable constraints and prior history.
Changing connectivity providers is different. An address allocation associated with an upstream provider’s aggregate may carry return or renumbering expectations when the network leaves that provider. Keeping the same prefix after such a change can affect routing aggregation. Changing the registry that administers a record would not, by itself, create a right to retain provider-based addresses after changing connectivity.
A transfer between holders is also different. When one organisation relinquishes a resource and another becomes the recognised holder, the allocation state changes substantively. Registry-service portability keeps the holder constant. It cannot serve as a disguised transfer or permit the destination registry to treat an existing allocation as a newly issued resource.
Geographic movement introduces another separate question. A holder may move its headquarters, infrastructure or customers without changing the resource or registry. A registry relationship might change while the network remains in the same location. Neither event necessarily entails the other.
These distinctions determine whether the proposed alternative remains narrow enough to analyse. A service handoff would have to leave the resource’s identity, holder, delegation history, allocation type and routing-related conditions intact. It could not turn provider-based space into provider-independent space, erase a return obligation or substitute a new policy merely because another institution had accepted the customer.
The authoritative record would consequently need to distinguish several relationships that an exclusive system could leave bundled together. It would identify the resource and current holder, record where the resource sat in the delegation chain, preserve the conditions governing it and state which registry was currently authorised to provide service. A port would change the last relationship without silently modifying the others.
The distinction between network operation and registry authority is equally important. Operators announce routes, maintain connectivity and troubleshoot failures. They may depend on accurate registration information and bear external costs when address policy affects routing. Those operational responsibilities do not give them authority to change the recognised holder or allocation status. Conversely, a registry’s authority over the record does not make it responsible for operating the holder’s routes.
Portability would not convert number resources into ordinary private property either. The historical guidelines treated registration and allocation as public coordination functions involving uniqueness, stewardship and continuing conditions. Replacing the servicing institution would not free a holder from those conditions or turn the common record into a private account controlled without external obligations.
The proposed benefit is correspondingly limited. A holder could leave an administrative relationship without demanding a new version of the underlying facts. The destination registry might differ in support, communication, reliability or institutional character, but it could not offer a more convenient allocation truth.
This restriction is not incidental. It is what separates plural service from rival allocation systems. Interchangeability could exist only within common validity rules strong enough to bind the outgoing provider, the destination provider and the holder.
Regional exclusivity answered real coordination problems
A fair counterfactual must give the regional model its strongest case.
The 1992 proposal confronted growth in a system whose central administrative capacity could not remain the sole point for every request. Regional delegation distributed work while retaining a legible hierarchy. One institution serving a defined region made it easier to identify who was responsible for receiving applications, applying policy and maintaining relevant records.
Local language and custom mattered because address administration involved communication, documentation and judgment, not merely copying identifiers into a list. Regional knowledge could assist applicants and support sub-allocation decisions made within the conditions known to the institution. RFC 1366 also connected the regional arrangement to fairness, giving it a public-administration rationale beyond convenience.
A defined regional institution could provide a venue for policy discussion among affected networks. That potential benefit should not be confused with proof of universal consent or equal power, but it remains a coherent reason for connecting service and community. Address policy affected shared resources and externalities, so administration could reasonably be designed around more than an individual customer relationship.
Routing aggregation made the stability case more demanding. Provider-based allocation and renumbering were responses to the need to limit fragmentation in global routing. If several registries could compete by loosening the conditions attached to addresses, the immediate customer and destination provider might obtain a benefit while other networks carried the routing consequences.
Regional exclusivity could reduce the number of organisational boundaries across which those constraints had to be enforced. A servicing registry responsible for a defined allocation domain would not need to negotiate every routine action with another provider seeking the same holder’s business. Responsibility would be easier to locate, even though the historical evidence does not measure whether this arrangement produced fewer errors or lower costs.
That qualification matters. The documents establish stated rationales and institutional structure. They do not provide comparative conflict rates, failure rates, service times or cost figures. It would therefore be inaccurate to claim that one registry per region was shown to prevent disputes, create trust or outperform a portable alternative. Its advantage in the record is an architectural argument about simplicity and responsibility, not an observed result against a tested competitor.
The same discipline applies to legitimacy. A registry can be regionally embedded and useful without every affected operator having authorised permanent dependence on it. Community support is not a count of individual consent. Participation in a discussion is not necessarily decision power. Legal membership does not automatically confer control of a board. Paying for service does not establish agreement with every policy. Continued use where no lateral alternative exists is not evidence that the user evaluated and rejected competing providers.
Nor does an opportunity to comment establish access to an enforceable remedy. A policy forum, an institutional membership structure, a parent appeal and judicial review are different channels with different powers. Their existence cannot be collapsed into a single claim that the community controlled the institution.
These distinctions do not negate the regional case. They prevent its operational rationale from being enlarged into an unsupported mandate claim. The record shows why regional administration was attractive and how it was organised. It does not supply the denominator needed to measure universal acceptance or compare the distribution of influence across communities.
A portable alternative would have had to retain the benefits that mattered while changing the service relationship. It could not simply replace regional administration with distant firms pursuing individual customers. It would need to preserve language access, policy consistency, fair allocation and attention to routing externalities, whether through the destination provider, a regional policy institution or the common authority above them.
That requirement makes the alternative more institutionally complex than a simple market analogy suggests. The choice was not between an arbitrary monopoly and unconstrained competition. It was between different ways of locating responsibility around an identifier system that could tolerate only one valid final state.
RFC 2050 clarified the limits of the existing remedies
RFC 2050, published in November 1996, described registration as serving public uniqueness and troubleshooting functions. It treated regional registration as the primary arrangement and placed resources within an administrative hierarchy. Its provisions help distinguish portability from remedies and movements already recognised by the system.
An appeal to a parent registry moved a dispute upward. It allowed a superior institution to review a decision made lower in the hierarchy. That could provide correction, but it did not replace the organisation that would continue servicing the holder. The remedy was vertical and decision-specific.
A holder seeking portability would be asking for something different: continued recognition of the same resource under the same common rules, but with another registry administering the relationship. The holder might agree with the applicable allocation policy while objecting to delay, communication, reliability or conduct at the service layer. Correcting one disputed decision and ending an ongoing service relationship are not equivalent remedies.
RFC 2050’s treatment of provider-based assignments draws another boundary. A network leaving a connectivity provider could be required to return addresses and renumber in support of aggregation. A registry port could not waive that obligation. If Registry B inherited the record, it would also inherit the resource’s status and the duty to enforce the same applicable constraint.
This is where a careless portability claim would become harmful. A destination registry might attract holders by promising that a service handoff changed the nature of the resource. If it could reclassify provider-based space, ignore a return requirement or treat a routing exception as an ordinary customer preference, service competition would become a route around common policy.
Party-to-party transfers provide a third distinction. RFC 2050 required registry approval for movement between parties. The provision demonstrates that resource state could change through a controlled act. It does not establish portability because the holder changes in a transfer and remains unchanged in a service port.
Appeal, renumbering and transfer therefore cannot be cited as early forms of lateral registry choice. Each addressed a separate relationship: review of a decision, use of addresses after a connectivity change, or movement between holders. None allowed the holder to retain the same resource while replacing the continuing registration provider.
By 1996, the hierarchy had a clearer internal logic. A resource entered through an allocation chain, relevant disputes could move to a parent and provider-based status could carry renumbering consequences. These arrangements aligned administrative responsibility with aggregation and delegation. They also left service exit outside the ordinary set of remedies.
The record does not reveal whether this omission caused material harm. There is no comparable dataset for fees, response times, denials, completed appeals, service interruptions or holder satisfaction. It cannot show whether a lateral option would have been frequently used or whether vertical review was inadequate in practice.
What RFC 2050 does show is that portability would have required more than permitting another office to edit a record. The destination would need to inherit the complete allocation history and apply the same substantive conditions. Otherwise, the holder could use a change of administrator to obtain a change that the common policy did not allow.
That requirement significantly narrows the possible field of competition. Registries could differ in the way they delivered service, but not in whether the same resource was simultaneously valid for different claimants or whether an inherited obligation existed. The historical evidence does not establish whether the remaining scope for differentiation would have justified the cost of building portability.
DNS made separability visible without proving the address case
The closest institutional comparison emerged in domain-name policy.
The US government’s Statement of Policy on the Management of Internet Names and Addresses, published on 5 June 1998, pursued registrar competition, equal access and shared registration. It also recorded concerns that non-portable registry arrangements could create lock-in and switching costs.
The consultation record included more than 430 initial comments comprising about 1,500 pages and more than 650 comments on the Green Paper. Those figures describe submissions at stages of consultation. They are not support percentages, user denominators or proof of consensus. The statement was US policy and said that it lacked force of law.
For this counterfactual, the comparison matters because the proposed DNS arrangement separated authoritative uniqueness from customer-facing registration service. Multiple registrars could serve customers while a shared system preserved a coherent answer about a domain’s registration. Competition did not require each registrar to maintain an incompatible authoritative claimant for the same name.
That made a similar separation imaginable for number resources. A common validity layer might preserve the holder, resource and allocation conditions while recognised registries delivered administrative service. Moving between registries would alter the provider relationship rather than create another identifier.
The comparison cannot establish feasibility. Domain names and routed address blocks do not produce the same externalities. Changing a registrar does not ordinarily determine whether the global routing system must carry another more-specific route. Address resources, particularly provider-based allocations, interact with topology, aggregation and renumbering policy.
Reverse DNS also binds an address record to delegated operational authority. An address port confined to customer records could leave a material dependency with the former registry. Moving that dependency would require coordination beyond the service relationship visible to the holder.
The White Paper therefore supports only a bounded inference. By 1998, coordinated uniqueness and plural registration service had been combined in public institutional design for a related identifier system. Permanent dependence on one service organisation was not synonymous with uniqueness across every Internet identifier context.
It does not follow that address registries could have copied the arrangement. The address case needed its own treatment of aggregation, delegation, allocation history and conflicting authority. DNS experience identifies a distinction that could be investigated; it does not provide the missing number-resource architecture.
The comparison also cannot establish historical motive. The different paths of names and numbers do not show that address policymakers rejected a tested, production-ready alternative. Different technical properties supplied legitimate reasons for caution. The materials do not record a completed portable-address design placed before decision-makers and consciously declined.
This is the point at which the counterfactual must resist retrospective neatness. The separation was institutionally visible in an adjacent field, but visibility is not deployability. The relevant question remains whether an address-specific handoff could preserve every condition that regional administration kept together.
The handoff would have concentrated the hardest authority above the providers
A portable arrangement would be tested not by an uncomplicated request but by a disputed one.
Consider a holder served by Registry A seeking to move to Registry B. The common system would need to establish that the request came from a representative authorised to act for the recorded holder, that Registry B was eligible to accept the relationship and that all existing conditions would follow the resource. It would also need to identify the exact point at which Registry A lost authority to submit further changes.
If both registries issued inconsistent instructions, recognition of each institution would not be enough to resolve the conflict. Accepting both would destroy coherent state. Always accepting the incumbent’s instruction would allow the source registry to obstruct exit. Always accepting the destination’s instruction would make a fraudulent request a means of administrative seizure.
Finality would therefore have to come from rules or an authority operating above the competing providers. That layer would authenticate the relevant actors, order conflicting instructions, determine when the handoff became effective and preserve the record required to reverse an error without creating simultaneous valid states.
This need changes the character of the alternative. Portability appears to decentralise the customer relationship, but it requires a shared mechanism with sufficient authority to constrain every provider. Registries could compete in support and service; they could not compete over which claimant was valid for the same resource.
Fraud makes the problem more difficult because organisational identity changes over time. Contacts become stale, staff leave, companies reorganise and credentials may be compromised. A destination provider might be presented with apparently valid evidence by a person lacking current authority. A source registry might identify a genuine inconsistency or might invoke verification concerns to delay an unwanted departure.
Common authentication standards would be necessary because the consequences of a weak check would extend beyond the destination provider. The holder, former registry and operators relying on the record could all be affected. If providers were free to reduce verification in order to attract incoming holders, competition could weaken the integrity of the common state.
The destination would also have to inherit unresolved matters. A port could not erase an allocation dispute, outstanding return condition or recorded transfer restriction. Otherwise, changing providers would offer a procedural escape from obligations that were supposed to remain common.
Failure presents a related but distinct challenge. Portability is attractive when the incumbent registry cannot continue serving holders, yet this is also the point at which ordinary handoff evidence may be unavailable. The source might not be able to authenticate requests, provide supporting records or relinquish authority through the normal route.
A common record independent of the provider could preserve resource identity, holder status and allocation history. But continuity would still require someone to determine that the source was unable to act and to authorise successor service. It would also require enough information outside the failed institution to validate holders and resolve competing recovery claims.
The historical material does not demonstrate how that recovery would work. It does not establish what documentation would reside in common custody, how a destination would verify records formerly maintained elsewhere or how bundled operational functions would continue. Portability cannot be credited with resilience merely because replacement is conceptually available. Replaceability becomes resilience only if authority, information and material dependencies survive the failed provider.
Reverse DNS illustrates the dependency problem. If the customer relationship moved while reverse-DNS authority remained under Registry A, exit would be incomplete. Registry A would still control a function associated with the resource. If that authority moved to Registry B, the transition would need to prevent an interval of overlapping or absent control.
A common infrastructure operator might retain the relevant parent function while registries submitted authenticated changes. That arrangement could reduce dependence on either service provider, but it would place additional power and responsibility in the common layer. The legitimacy question would shift rather than disappear.
Later security assertions tied to resource state would create the same demand for orderly authority. The period ends in 2005, so later certification arrangements cannot be treated as a historical solution. The prospective lesson is narrower: any durable portability model would eventually need to end the former provider’s authority, establish the destination’s authority and avoid a gap or overlap.
Technical authentication alone could not decide every dispute. It might verify that an instruction came from a particular key, but it could not determine a contested corporate succession, decide which representative had legal authority after an organisational change or reconcile incompatible court orders without prior governance rules.
Cross-border service would expose that last problem directly. A holder, policy region, network footprint and destination registry might be connected to different jurisdictions. If one legal authority instructed the source registry to freeze a resource while another instructed the destination to complete a port, the common record could not give effect to incompatible outcomes.
Keeping governing law attached to the resource’s policy domain might preserve continuity but reduce the jurisdictional choice offered by portability. Attaching it to the destination could make a service port alter legal exposure and invite attempts to escape inconvenient obligations. A shared contractual framework could allocate duties among registries, but it could not erase national law.
These difficulties do not prove that portable service was impossible. They identify where its governing authority would reside. A system with interchangeable providers would still depend on a non-interchangeable rule of finality capable of deciding contested transitions, preserving inherited conditions and coordinating bundled functions.
That shared authority could become more remote than the regional institution it made replaceable. Its quality, legitimacy and vulnerability would therefore be central to the comparison. Portability reallocates institutional dependence; it does not abolish it.
Aggregation limits what honest portability could promise
CIDR, provider-based assignment and renumbering supply the strongest technical warning against an expansive portability claim.
A network receiving addresses within a connectivity provider’s aggregate may create a routing consequence if it retains and announces the same addresses after leaving that provider. The holder values continuity, but the resulting route can impose a burden outside the transaction. RFC 2050’s return and renumbering expectations were connected to this aggregation concern.
Registry-service portability cannot honestly promise that a prefix will follow a holder through every topology or connectivity change. That would combine administrative exit with provider-independent routing and obscure the externality that aggregation policy was designed to manage.
The narrow model leaves routing status unchanged. Registry B administers the same allocation type and the same conditions that applied under Registry A. If the holder later changes connectivity providers, the applicable routing and renumbering rules are considered separately. The registry port neither creates nor removes a right to retain the prefix.
Enforcing this boundary would require common state precise enough to prevent reclassification during a handoff. The destination could not market portability as a route to more permissive treatment of provider-based space. An inherited return obligation would remain visible and binding.
That restriction may leave less room for competition than the language of interchangeability implies. Registries could differ in communication, administrative support, responsiveness or institutional form, but they could not offer different versions of the resource’s allocation status. The more that must remain common for routing stability, the smaller the contestable service layer may become.
A regional model had a credible advantage in this respect. A limited set of institutions with defined responsibility could connect address administration to the communities and routing conditions affected by the decisions. A portable system would have to reproduce the same discipline across provider boundaries.
The fixed evidence does not measure the routing effect of a narrowly defined registry-service handoff. It also does not demonstrate that the identity of the servicing registry was itself necessary to preserve aggregation. The justified conclusion is conditional: if changing registries allowed a holder to evade topology-related obligations, portability would export costs and fail its own narrow definition.
This condition also limits the later portability claim. An ability to move IP addresses or ASNs between registries cannot mean that the destination may duplicate the resource, alter its holder, change its routing properties or disregard allocation constraints. To remain a service port, the movement must be administrative and governed.
Recognition policy reinforced durable regional institutions
ICP-2, accepted on 4 June 2001, placed regional service, community support, stable operation and coordination at the centre of recognising new Regional Internet Registries. It described three existing RIRs covering the world, anticipated institutions for Africa and Latin America and expected the number of RIRs to remain small.
For this inquiry, the relevant point is the institutional form that ICP-2 recognised. It contemplated durable regional organisations, not interchangeable providers available to individual holders. Another region could establish a registry under the stated criteria, but that kind of institutional entry did not give a holder lateral exit from its continuing service relationship.
ICP-2 was written by incumbent RIRs with the Address Supporting Organization and accepted by ICANN. That provenance does not invalidate the criteria or the operational knowledge behind them. It does mean that the document records how the established architecture defined legitimate expansion. It cannot independently demonstrate that the resulting number of institutions was optimal or that every operator authorised permanent regional attachment.
Community support is particularly easy to overread. It can be relevant to recognition without functioning as a referendum of all resource holders. Participation in a policy discussion may offer voice while leaving formal decision power elsewhere. Legal membership, voting rights, board control and access to an enforceable remedy are separate institutional relationships.
Payment and service use are separate again. A holder may pay fees because payment is required to maintain a recognised service relationship. That act does not establish consent to every policy or to the absence of alternatives. Continued use cannot be treated as a revealed preference for exclusivity where lateral portability was not generally available.
Operator responsibility also should not be mistaken for registry authority. Operators bear the practical work of announcing routes, maintaining network reliability and troubleshooting problems. Their exposure to routing consequences gives them an important interest in address policy, but it does not itself authorise them to modify allocation state. The registry’s authority to record changes likewise does not make it the operator of the routes.
By keeping these relationships distinct, the legitimacy question becomes more precise. ICP-2 shows that the recognised model valued regional community support, operational stability and coordination. It does not quantify how influence was distributed, whether every entity could obtain an effective remedy or whether individual holders preferred the absence of lateral choice.
A portable model would have required a recognition category that ICP-2 did not describe. The common system would need to determine which organisations could provide service, which policy domains they could serve, what obligations they inherited and how their authority could be withdrawn. Regional institutions might continue to develop policy while service providers administered individual records, or some organisations might perform both roles.
Such separation could preserve regional policy participation while permitting service exit. It could also make responsibility harder to trace. A provider outside the policy community might administer a resource whose constraints were set elsewhere. If it had no discretion, service differences might be modest; if it had broad discretion, policy inconsistency and arbitrage could follow.
The record does not show that this alternative recognition structure was specified, supported or tested. By 2001, the accepted framework reinforced a small population of durable regional institutions. That is evidence of the architecture in force, not a comparative verdict on every untried arrangement.
Exit and voice would have addressed different failures
The governance argument for portability rests on exit, but the mechanism should not be credited with outcomes that were never observed.
Voice gives an affected party ways to contest or influence a decision: participation in policy discussion, requests for explanation, appeals within the hierarchy, collective action, or political and legal challenge. These routes seek to change a decision or the institution’s conduct while the entity remains within the relationship.
Service exit would replace the administrator of an unchanged record. It would be relevant when the complaint concerned the institution providing service rather than the validity of the common allocation rule. A holder might seek another provider because of communication, delay, reliability, treatment or concern about institutional continuity.
The destination could not legitimately override a common-policy denial merely to win the holder’s business. If an allocation request failed because the shared rules did not permit it, switching providers could not erase that determination. Otherwise, exit would turn common policy into an optional constraint.
A parent-registry appeal and a service port would therefore solve different problems. An appeal can correct a decision while leaving the continuing relationship intact. A port can end the relationship while leaving the substantive rule intact. Neither is a complete substitute for the other.
The possibility of exit might affect provider conduct, but the historical evidence does not measure that effect. It contains no switching experience from which to estimate changes in responsiveness, price, reliability or treatment. It also lacks comparable evidence showing how much variation existed in service quality.
Portability might provide a continuity route if a registry became unable to serve holders, but that benefit would depend on the recovery arrangements discussed earlier. Formal choice without transferable records, successor capacity and authoritative finality would not produce dependable continuity.
Voice has its own limits. The opportunity to attend a meeting or submit a comment does not establish equal decision power or timely redress. Community support does not reveal whether a particular holder could reverse a contested action. An appeal can provide review without offering a way to end the service relationship.
The most plausible institutional objective would therefore combine regional voice with bounded service exit. Common policy could remain subject to collective governance while the administration of an individual record became replaceable under shared rules. That combination is a normative design possibility, not an observed historical outcome.
Sharon Gillett and Mitchell Kapor’s 1997 analysis, The Self-governing Internet: Coordination by Design, helps frame the boundary. They distinguished the decentralised bulk of Internet operation from the exceptional coordination required for identifiers and asked whether names and address allocation could become more like the surrounding Internet.
Their analysis did not provide a portable address-registry protocol or test one in operation. Its contribution was to prevent exceptional coordination from becoming an answer to every organisational question. Identifier validity required concentration or tightly governed federation, but that fact did not by itself determine how all administrative service had to be delivered.
The counterfactual follows that distinction cautiously. Non-duplication, delegation validity and final conflict resolution belong in the coordinated layer. Support and some administrative functions might be provided by more than one institution. The difficult boundary is the handoff that allows a provider to change without allowing the underlying facts to diverge.
The 2025 proposal makes the normative claim explicit
Heng Lu’s 17 September 2025 proposal, On Portability of Number Resources and the ICP-2 Revision, argues for an unconditional ability to move IP addresses or ASNs between registries, presenting exit as a mechanism for resilience and accountability.
The date limits what the proposal can establish. It does not prove that designers in 1992 intended portability, that operators between 1992 and 2005 demanded it, or that the required safeguards could have been deployed during the period. It is a later normative argument, not evidence of an earlier operational capability.
Its value is that it states the institutional concern directly. A resource holder may need a way to leave a registry without surrendering the resource merely because the service relationship changes. That claim separates dependence on a provider from the need for coherent number-resource state.
The historical evidence gives the question legitimacy without validating the answer. RFC 1366’s central fallback shows that the intended regional contact point was not the only imaginable administrative point in every circumstance. The DNS policy comparison shows that plural service and common uniqueness were combined in a related identifier design. Neither establishes that number-resource portability could preserve address-specific constraints.
The term “unconditional” must therefore be interpreted with care. It cannot mean that a destination registry may recognise a duplicate claimant, erase an unresolved dispute, change the holder without an approved transfer, disregard provider-based conditions or leave reverse-DNS and security authority ambiguous. It cannot make incompatible legal commands simultaneously valid.
A defensible normative version would attach the entitlement to a governed service handoff. A legitimate holder would not be kept with an incumbent merely to preserve that provider’s customer relationship, provided the common system could authenticate the request, carry every obligation forward and complete all material transitions safely.
Even that narrower objective remains unverified. The evidence does not establish whether recognised registries would accept mandatory incoming ports, how contested requests would be adjudicated, how much the system would cost, whether providers would offer meaningful differences or how often holders would use the mechanism.
The proposal is therefore most useful as a test directed at the inherited architecture. It identifies a form of exit that the regional hierarchy did not centre. It also reveals that credible exit would depend on a powerful common layer capable of preserving finality when providers disagreed.
What the evidence can and cannot carry
The historical material is sufficient to distinguish technical uniqueness from permanent service attachment. It is not sufficient to estimate the comparative performance of the two institutional arrangements.
For the regional model, the sources establish stated reasons: fair sub-allocation, language and custom, administrative scale, regional service, stability and coordination. They do not supply comparable measurements showing that regional exclusivity reduced errors, prevented failures, created trust, lowered costs or produced superior service.
For portability, the sources establish a mechanism worth examining: an unchanged holder might replace a servicing registry while common state and inherited conditions remained intact. They do not establish that exit would improve responsiveness, reduce fees, prevent discrimination, increase resilience or produce better governance.
There is no consistent 1992–2005 dataset comparing prices, processing times, denials, appeals, interruptions, holder satisfaction or switching. Consultation counts from the NTIA record cannot substitute for those measures, and they do not reveal support for address-registry portability.
The material contains no production specification for a lateral handoff, no interoperability trial, no cost comparison and no failure exercise. It provides no demonstrated resolution for simultaneous claims, failed-provider recovery, reverse-DNS continuity, later security transitions or incompatible legal commands.
Those are not interchangeable omissions. Service data would be needed to estimate whether choice addressed a material problem. A handoff specification would be needed to judge whether the change could preserve authoritative state. A cost comparison would be needed to determine whether portability’s common infrastructure and dispute mechanisms were proportionate. Failure exercises would be needed before replaceability could be treated as resilience.
The absence of these records does not prove that the alternative would have failed. It also cannot be used as evidence that the alternative would have worked. The disciplined position is uncertainty about operational feasibility and outcomes, combined with a narrower logical finding about what uniqueness alone entails.
The regional system’s continued development through 2005 does not fill the comparative gap. Institutional continuation shows that the architecture persisted; without the relevant denominator and outcome evidence, it cannot demonstrate that every bundled function was necessary or that an alternative would have been worse.
Likewise, later advocacy cannot supply the missing historical test. A clearer normative objective in 2025 does not reveal readiness, acceptance or likely performance during the earlier period.
The strongest counterevidence remains intact. Regional administration could improve access to language and custom and provide a legible location for responsibility. Provider-based allocation and renumbering reflected a genuine aggregation concern. DNS supplied only an analogy because routed prefixes carry different externalities. Exit would be meaningful only if common state prevented duplication, obstruction and selective rejection.
Together, those constraints explain why the counterfactual cannot be reduced either to a missed opportunity or to a technical impossibility. The unresolved question concerns the quality of the institutional boundary: whether enough service could have been made replaceable without weakening the coordinated functions on which everyone depended.
A bounded judgment
The evidence rules out a claim of necessity based on uniqueness alone. A globally coherent allocation record requires authoritative state, but that requirement does not logically identify the organisation that must provide every holder’s administrative service forever.
The evidence also rules out confidence in the alternative. No demonstrated arrangement from 1992–2005 shows that a lateral service handoff could preserve allocation conditions, final authority and all material operational dependencies under dispute or failure.
The decisive criterion is not whether two registries can read the same record. It is whether a contested change of provider can end with one authoritative state, an intact chain of obligations and no residual control left ambiguously between the institutions. A proposal that cannot meet that criterion offers nominal choice at the cost of coordination. A proposal that can meet it would still need comparative evidence before its safety or benefit could be claimed.
Regional exclusivity addressed identifiable problems and should be judged on those merits, not recast as proof of universal consent. Portability identifies a legitimate accountability question and should be judged by the handoff it can demonstrate, not by the attractiveness of exit in the abstract.
The counterfactual therefore leaves one productive burden of proof. Permanent attachment cannot be defended merely by invoking uniqueness, while replaceable service cannot be defended merely by imagining competition. The unresolved institutional frontier lies in proving that authority, obligations and operational control can all survive the provider that currently holds them.