Summary

  • A global number policy is not a single ICANN vote. It is a staged consent chain: the same policy must move through the five RIR communities, be assembled into common text, be reviewed by the ASO Address Council, be referred through the NRO Executive Council, and then be accepted, rejected, sent back or allowed to take effect through ICANN Board inaction.
  • The post-exhaustion IPv4 policy shows the chain at work because earlier recovered-address proposals failed when versions diverged or regional consent broke, while the 2011 version succeeded after identical regional adoption and Board ratification in 2012.
  • The procedure protects regional autonomy and IANA neutrality, but it also disperses accountability. Each gate should publish the reason for approval, delay, return, extension or rejection so the public can see whether a proposal failed on substance, process, wording, timing or institutional caution.

A global policy is a chain, not a moment

A global number policy does not become real because one meeting decides it should be real. It becomes real because a long chain of institutions accepts the same act, in the same field, for the same limited purpose. The RIR communities must agree. The ASO Address Council must satisfy itself that the regional steps were properly followed and that significant viewpoints were considered. The NRO Executive Council must refer the coordinated text. The ICANN Board must decide whether to accept, reject, request changes or let the policy take effect by not acting within the defined period.

Public Technical Identifiers then carries out the IANA numbering action the policy requires.

That chain is awkward by design. It reflects the political fact that IANA number allocations are global while number policy authority is regionally rooted. A rule about recovered IPv4 space, ASN blocks or IPv6 allocations must be compatible across all regions because IANA cannot run five conflicting global allocation rules. Yet a rule cannot be imposed from ICANN downward without undercutting the regional policy forums that give RIR policy its legitimacy.

The ASO global policy page states the boundary directly: global policies govern how PTI, as IANA functions operator, issues Internet number resources to the RIRs, while the RIRs distribute those resources under their own regional policies. The NRO global policy page uses the same functional division. Global policy is the rule for the top allocation point. Regional policy is the rule for the distribution that follows.

This division creates a useful but often misunderstood form of restraint. ICANN cannot simply write a number policy and order the RIRs to accept it. A single RIR cannot write a global IANA rule for the others. The ASO AC cannot turn process review into a substitute legislature. The NRO EC cannot treat executive coordination as a private veto over regional communities. The ICANN Board cannot reject a policy without supplying concerns and, after a second rejection, moving toward mediation. Each entity holds a limited tool, not a complete mandate.

The problem is that limited tools can still create delay. A proposal can stall in one RIR. It can pass in different forms and fail the common-text test. It can reach the ASO AC and be returned for further regional review. The ICANN Board can request changes. The Board can reject once, reject a resubmitted version by supermajority, or simply allow the clock to run. Every stage is defensible in isolation. The governance question is whether the public can trace which stage did what, why it did it, and who must answer for the resulting delay.

The post-exhaustion IPv4 proposals are the cleanest way to see the mechanism. They were not abstract constitutional exercises. They dealt with a concrete problem: what IANA should do with IPv4 space returned after the central free pool had already run out. The answer mattered because any recovered pool would be small, politically sensitive and economically visible. Equal distribution, need-based distribution, mandatory return and voluntary return each favored different theories of fairness.

The eventual 2012 policy worked because the system forced the proposal through all gates. Earlier versions failed because the same system refused to treat partial agreement as global consent. That is not bureaucratic trivia. It is the core accountability lesson. A global policy chain has value only if its refusal points are as legible as its final approvals.

The IANA action defines the global field

The first discipline is scope. A proposal is global only when it requires an IANA or ICANN-related action beyond one region. The ICANN background report for the 2011 post-exhaustion IPv4 proposal explained that global Internet number resource policies are policies agreed by all RIRs under their own procedures and by ICANN, and that require specific action or outcome by IANA or another external ICANN-related body. That definition keeps ordinary regional disputes from being pulled into global machinery merely because several regions care about them.

The post-exhaustion IPv4 case fits the definition. The question was not whether one RIR should change its transfer market, waiting list, need test or member rule. The question was what the IANA functions operator would do if address blocks came back after the central IPv4 pool had already been exhausted. The final policy, published by ICANN as the Global Policy for Post Exhaustion IPv4 Allocation Mechanisms by the IANA, established a recovered IPv4 pool, described what would go into it, set an activation trigger, defined allocation periods, and used a formula that divided the pool into equal allocation units for the RIRs.

That is the sort of decision that cannot sensibly be left to one region. If returned space enters an IANA pool, every RIR has an interest in the trigger, allocation unit, frequency and transparency of distribution. If IANA uses need alone, regions with faster exhaustion may receive more. If IANA uses equal distribution, every RIR receives the same unit even when regional need differs. If returns are mandatory, the rule changes each RIR's control over recovered assets. If returns are voluntary, the recovered pool may remain small. These are global tradeoffs.

The global path therefore begins with a narrow question: what must the central number function do? The more clearly that question is framed, the less likely the process is to become a forum for every regional grievance attached to the same resource. In the IPv4 case, the final policy's strength was its narrowness. It did not decide the entire economics of IPv4 scarcity. It did not create a global transfer market. It did not tell RIRs how to allocate scarce space to their members. It told IANA how to hold and distribute returned fragments under a specific post-exhaustion formula.

That narrowness also helped accountability. If the policy failed, the public could ask whether the failure came from the recovered-pool trigger, the equality formula, the treatment of returned space, the minimum allocation unit or the regional adoption sequence. A broad scarcity manifesto would have created too many reasons for refusal. A narrow IANA action allowed each institution to state a more precise concern.

Future global proposals need the same discipline. Before debate begins, the proposer should identify the exact IANA action, the operational consequence for each RIR, the regional matters left untouched, and the reason no region can solve the problem alone. This does not depoliticize the decision. It makes the politics auditable.

Five regional gates protect legitimacy and create veto power

The regional gate is the strongest part of the global policy design. The ASO Global PDP description says a proposed global policy may be submitted to an RIR policy forum or directly to the ASO Address Council, but it must be made known across all regional policy forums. The proposer has a duty to help communities understand deliberations elsewhere. If the advocate cannot travel to a public policy meeting, the RIR is to appoint someone to present the proposal. Common elements are then documented as each region considers the text.

This design prevents a global policy from being manufactured in a single room. It forces a rule that affects all RIRs to be tested against different membership structures, languages, meeting cultures, legal environments, address economies and historical expectations. A policy that seems neutral in one region may expose a burden in another. A need-based formula may be attractive to a region with acute scarcity and suspect to a region concerned about incentive effects. A mandatory return clause may look administratively clean and still be politically unacceptable where returned space is tied to local recovery practice.

The cost is that every region receives a practical veto. The veto may not be called that. It may appear as no consensus, abandonment, withdrawal, delayed final call, altered wording or board non-ratification after regional discussion. But if all five regions must adopt an identical common text, one region's refusal blocks global policy. That is not a flaw hidden in the system. It is the system.

The 2010 post-exhaustion proposal shows the point. ICANN's background report stated that the proposal had been adopted in ARIN, remained under discussion in LACNIC and AFRINIC, but had been abandoned in APNIC and withdrawn in RIPE. Because it did not reach consensus in all five regions, it was no longer eligible to become global policy unless negative decisions were reconsidered. The reason was not that ICANN opposed the idea. The reason was that the regional chain was broken.

That regional veto is legitimate only when the reason is visible. If a proposal fails in APNIC because the community prefers a different allocation formula, the record should show that. If it fails in RIPE because the text does not match local consensus, the record should show that. If it stalls because translation, meeting cadence or presenter availability delays debate, the record should show that too. Otherwise a distributed legitimacy check can become a fog of institutional silence.

The five-gate structure also changes advocacy. A proposer must explain the rule repeatedly in different venues and adapt to different objections without losing common text. That is a high burden. It favors actors with time, travel support, procedural fluency and cross-regional relationships. The ASO PDP attempts to reduce this burden by requiring awareness across regions and allowing an RIR to appoint a presenter if the advocate cannot travel. But the practical burden remains real.

For that reason, the legitimacy of a global policy should not be measured only by final ratification. It should be measured by the quality of the regional record: who was notified, what objections appeared, how text changed, whether late regions received the same information as early regions, and whether each adoption decision explained the same text in comparable terms.

Common text is the quiet control point

The most technical gate may be the most political one: common text. The ASO process recognizes that outcomes in regional discussions may differ in language and detail. RIR staff are expected to work with one another and with the proposer to document common elements. The common text is then ratified by each RIR, by methods of its own choosing. That text becomes the proposed global policy forwarded to the ASO AC.

This sounds clerical. It is not. Common text decides whether five regional decisions are truly the same decision. If one region adopts need-based allocation, another adopts equal allocation and a third adds a return obligation, the system cannot pretend that a global policy exists. The policy may share a theme, but IANA cannot execute a theme. It needs a rule.

The 2009 and 2010 recovered IPv4 history shows how decisive common text can be. The 2011 ICANN background report noted that a previous 2009 proposal was abandoned by the NRO EC in view of version differences across the RIRs. It also noted that the 2010 version lost viability after APNIC and RIPE moved away from it. In both cases, the failure was not merely a lack of enthusiasm. It was a failure to produce an identical, globally ratifiable instruction.

This is where responsibility can become blurred. A regional community may say it approved the principle. Another may say it approved a modified form. Staff may describe common elements. The NRO EC may decline to proceed because the versions differ. The ASO AC may never receive a clean text. To outsiders, that can look like a proposal died somewhere in coordination. In fact, it died because common text is an approval condition.

The remedy is not to weaken the common-text rule. The remedy is to make it explicit. Every global proposal should maintain a public comparison table showing live text in each region, material differences, decision status and the reason each difference matters for IANA execution. The table should distinguish editorial variation from substantive variation. It should mark when a proposal has a common core but not a common operative clause. It should also show who is responsible for reconciling language and when the reconciled text returns to each region.

Common text also protects minority regions. Without it, a large or fast-moving region could say that broad global sentiment supports a rule even though another region adopted a materially different safeguard. Requiring identical text gives each region a concrete document to approve. It prevents global policy from being assembled through paraphrase.

At the same time, the common-text gate should not become a hidden drafting veto. If staff coordination changes the substance, the change should go back to the regional public forum. If a difference is said to be editorial, the reason should be published. If translation across languages creates ambiguity, the operative English version and any authoritative local-language explanations should be compared openly. The quiet control point should not remain quiet when it decides whether global policy exists.

The 2009 failure shows how a proposal can die before ICANN

The first recovered IPv4 effort did not fail because the ICANN Board rejected it after a dramatic debate. It failed earlier, inside the multi-region chain. The ASO proposals page lists the 2009 proposal, "Global Policy Proposal for the Allocation of IPv4 Blocks to Regional Internet Registries," as abandoned because the same version was not adopted by all five RIR communities. The ICANN background report for the later 2011 proposal explains that the 2009 effort was abandoned by the NRO EC in view of version differences across the RIRs.

That is a useful failure. It shows that the global policy system can stop a proposal before it reaches the Board when regional approval is not the same approval. It also shows why the record must be detailed. To a resource holder trying to understand why recovered space has no global rule, "version differences" is not enough. The public needs to know which clauses differed, which regions accepted which version, whether the difference concerned mandatory return, allocation basis, activation trigger, timing or another issue, and whether the difference was considered curable.

The 2009 failure also shows that the NRO EC's role is not only ceremonial. If the EC is expected to handle or abandon a proposal when versions differ, then the EC is exercising a gate that affects whether the public ever sees an ICANN Board decision. That gate may be necessary. A CEO-level body should not send contradictory regional texts to the Board as if they were one global policy. But the decision not to advance a proposal should be explained as a procedural and substantive finding, not left as institutional shorthand.

This matters because failure before ICANN can be politically convenient. No Board vote means no ICANN rejection. No final ASO AC referral means no formal Board concerns. No single regional veto may be publicly named if several versions diverged. Everyone can say the process did not produce common text. That may be true, but accountability requires more than truth at a high level. It requires a map of what prevented common text.

A failed global policy should therefore receive a closing note. The note should identify the proposal, the versions, the regional statuses, the unsolved differences, the decision body that closed or suspended the file, and the options for revival. If revival requires a new proposal, say so. If a region could reconsider a negative result, say so. If differences are too deep, say so. The point is not to shame any region. The point is to let the public distinguish principled refusal from procedural exhaustion.

The 2009 recovered-space case became more than a failure because later proposals learned from it. The system did not abandon the problem. It abandoned a form that could not command identical regional consent. That is exactly how the chain should work, provided the learning is visible.

The 2010 failure shows that a regional no is a global event

The 2010 proposal is even more revealing because the record shows a split path. ICANN's background report said the proposal was adopted in ARIN, remained under discussion in LACNIC and AFRINIC, but was abandoned in APNIC and withdrawn in RIPE. It then drew the institutional conclusion: because consensus was not reached in all five regions, the proposal was no longer eligible to become a global policy unless a negative verdict was reconsidered.

This is the regional veto in practical form. ARIN's adoption did not carry the proposal. Continued discussion in two regions did not keep it alive once two other regions had moved away. ICANN did not rescue it. The NRO EC and ASO AC could not treat four partial states as a global mandate. A single global rule requires all five.

That rule protects the legitimacy of the IANA function. If IANA is to distribute recovered IPv4 space under a global formula, every region must be able to say that its policy forum accepted that formula. Without that, IANA would be implementing a policy that some regional communities had not accepted. The fact that IPv4 scarcity is global does not erase regional consent.

The same rule can frustrate actors who want a quick answer. IPv4 exhaustion did not wait for institutional neatness. Market pressure was already rising. Returned address space, however small, had economic meaning. A need-based formula could have seemed urgent to networks with less remaining stock. An equal formula could have seemed fairer to regions wary of strategic depletion claims. Every delay had distributional effect.

That is why a regional no should be treated as a global event. When APNIC abandoned the 2010 approach in favor of a different proposal, that was not merely an APNIC meeting outcome. It changed the likelihood that IANA would receive any recovered-pool rule through that path. When RIPE withdrew its version, that too altered the global timeline. The public record should help readers understand that a regional decision can reshape the central allocation function.

The 2010 episode also shows why proposals should not be evaluated only by final success. A failed proposal can reveal the design preferences that make success possible later. The APNIC preference for a new third proposal was part of the movement toward the 2011 formula. A no can be constructive when it is attached to a better alternative. It is destructive when it leaves no evidence, no reason and no route forward.

Future global policy pages should therefore treat regional negative decisions as first-class records. A status table should not merely mark "abandoned" or "withdrawn." It should capture the controlling reason, the alternative under discussion and the effect on global eligibility. If the proposal can be revived only through regional reconsideration, that condition should be stated in plain terms. The user of the policy system should not need to reconstruct the chain from scattered regional archives.

The 2011 proposal worked because it narrowed the bargain

The accepted post-exhaustion IPv4 proposal succeeded because it solved enough of the prior dispute without trying to solve everything. It did not require mandatory return of all recoverable space. It did not allocate based on individual regional need once the pool became active. It created a recovered pool, waited until a defined exhaustion trigger, and distributed one allocation unit to each RIR in a six-month period when the pool was large enough. The allocation unit was one fifth of the recovered pool rounded down to a CIDR boundary, with a minimum size of /24.

That design was politically and administratively significant. Equal allocation avoided a recurring fight over regional need. A defined period avoided ad hoc distribution. A minimum unit avoided meaningless allocations. Voluntary or otherwise returned space could enter the pool without requiring the global policy to adjudicate every regional return incentive. The rule was not a full theory of IPv4 justice. It was a limited central mechanism for a small and uncertain pool.

ICANN's 2011 background report shows how the proposal moved. The ASO AC recognized the proposal as satisfying formal requirements as a candidate for global policy on 3 February 2011. The proposal was then introduced and advanced across the RIRs. The report stated that it had been adopted in all five regions and would be handled by the NRO EC and ASO AC before submission to the ICANN Board. The final ICANN policy page states that the policy was ratified on 6 May 2012.

This path shows the procedure at its best. A global problem was identified. Multiple drafts tested competing formulas. Regional communities refused versions they could not accept. A narrower version emerged. Each region adopted it. The ASO AC and NRO EC handled the transition from regional text to global proposal. The Board ratified it. IANA received an executable rule.

The success also shows why time matters. The proposal was not born at the Board. It passed through months of regional introduction, public forum discussion, final calls and adoption decisions. Those dates matter because they show where delay was unavoidable and where it was institutional. A policy that requires five public communities cannot move at the speed of a management memo. But the record should show when waiting served deliberation and when waiting simply reflected queue position or unclear ownership.

The 2011 proposal's strongest lesson is that distributed vetoes can produce better text when they are tied to reasons. If APNIC's abandonment of the 2010 version helped move the system toward a more acceptable third proposal, then the veto was not just obstruction. It was part of design correction. But that benefit depends on visible reasoning. A veto without reasons teaches nothing.

The final policy also demonstrates the proper relationship between global and regional rules. It did not tell RIRs how to distribute the units they received. It stopped at IANA allocation. That restraint was likely central to its acceptance. Global policy should be heavy only where the global function is heavy. Beyond that, regional communities retain their own distribution choices.

The ASO AC is a process auditor with real leverage

The ASO Address Council's role is easy to underestimate because it is not described as a legislature. In the global policy process, it reviews whether the RIRs reached common agreement and common text, and whether significant viewpoints of interested parties were adequately considered. Within sixty days after the NRO EC advises that all regions adopted the proposal, the ASO AC must pass it to ICANN, advise the NRO EC that further regional review is needed, or request more time.

That is a process role, but process roles can carry power. If the ASO AC concludes that significant viewpoints were not adequately considered, a proposal that has regional adoption can still be sent back. If it requests more time, the Board does not receive the proposal yet. If it passes the proposal, it tells ICANN that the regional chain is procedurally ready. Each choice changes the timeline and the public meaning of the proposal.

The ASO AC should therefore be judged by the quality of its process findings. It should not restate every regional debate. It should identify the evidence it reviewed: regional adoption dates, final texts, last-call records, unresolved objections, presentation history, translation or access issues, and any material divergence among versions. If it finds that the proposal can proceed, it should say why the significant viewpoints threshold is satisfied. If it sends the proposal back, it should state which viewpoint or regional step needs further attention.

This is not a call for the ASO AC to second-guess regional policy substance. The point is to distinguish substance from process. A region may adopt a rule that some actors dislike. That is not itself a process defect. A process defect appears when material objections were not visible, when the common text did not match regional approvals, when a region did not complete its own required steps, or when the record does not show that affected viewpoints were meaningfully considered.

The Council's leverage is especially significant because global proposals are rare. The ASO current policy page lists only three current global policies: IPv6 blocks to RIRs, ASN blocks to RIRs, and post-exhaustion IPv4 allocation. Because the field is small, each ASO AC action sets precedent for how strict the process check will be next time. A loose review makes future regional approval easier to package. A disciplined review makes future proposals more expensive to document but more credible.

The ASO AC also sits between public community and institutional executive layers. Its members are the NRO Number Council: three from each RIR region, with elected and appointed components. That composition gives it a different legitimacy base from the NRO EC. The EC represents the RIR institutions. The ASO AC carries policy-community representation. A global policy chain needs both, but it should not let one speak as the other.

The clean rule is this: the ASO AC should make process adequacy visible, not merely transmit. Its review is the public's best check on whether a global policy is truly the product of five regional policy communities rather than five institutional endorsements.

The NRO EC is a referral gate, not a private legislature

The NRO Executive Council sits in a sensitive position. Under the NRO MoU, it represents the NRO and its suborganizations, can commit RIR resources when unanimous agreement is reached, and has a role in ratifying or rejecting proposed global IP number resource policies based on open and transparent procedures ratified by the regional address policy fora. The NRO global policy page says the EC refers the coordinated global policy proposal to the ASO AC.

The EC's involvement is necessary because the NRO is the coordinating body for the five RIRs. Someone must confirm that the RIR institutions are aligned on referral, that a common policy file exists, and that the proposal is ready to enter the ASO AC and ICANN stages. But the EC should not become a private legislature above the regional communities. Its authority should be tied to coordination, common text, resource commitment and procedural readiness.

This distinction matters because the EC is a CEO-level body. The NRO EC page says it consists of one person from each RIR, appointed by each RIR board, and acts only by consensus from all five regions. That consensus requirement is a restraint, but it is also a veto. If all five EC members must agree, then one EC member can stop a referral even after regional communities have done substantial work, depending on the procedural setting.

That may be necessary where versions differ or the common text is defective. It is dangerous if the reason is not public. A CEO-level refusal can look like institutional risk management rather than community policy judgment. If the EC finds that a proposal cannot advance, it should identify whether the barrier is lack of identical text, incomplete regional adoption, unresolved process appeal, resource commitment, legal issue or another defined reason. The public should not have to infer the cause from silence.

The EC also carries cost and implementation implications. A global policy may require IANA action, but the RIRs may need staff coordination, public communications, statistical reporting, engineering changes, member notices or joint legal review. The EC is the layer most likely to see those operational costs across all registries. That knowledge is valuable. It can prevent a policy from being framed as costless when it is not.

But cost awareness should not be confused with policy veto. If the EC believes implementation costs are material, it should publish a cost and readiness statement for the ASO AC and the Board. If a policy imposes burdens unevenly across regions, the record should say so. If a cost can be managed by timetable, phased start or reporting limit, that should be visible. The EC's executive knowledge should enrich the public record, not replace it.

The 2009 recovered-space failure shows why this discipline is needed. The proposal was abandoned by the NRO EC in view of regional version differences. That may have been a correct application of the common-text rule. The governance lesson is that any such decision should leave enough detail for later communities to understand what must be changed.

The ICANN Board has a narrow but consequential check

After the ASO AC forwards a proposal, the ICANN Board's role is not passive. The Global PDP gives the Board sixty days to accept by simple majority, reject by two-thirds supermajority, request changes by simple majority or take no action. If the Board does nothing within the window, the proposal is deemed accepted and becomes global policy. If the Board rejects, it must deliver a statement of concerns, with attention to significant viewpoints that were not adequately considered. If a resubmitted proposal is rejected a second time by supermajority, the RIRs or ICANN must refer the matter to mediation under an agreed procedure.

This structure gives the Board a real check while limiting its ability to dominate. The Board can ask questions and consult with the ASO AC, the RIRs acting collectively through the NRO and other parties it considers appropriate. But rejection carries a high threshold and a duty to explain. A request for changes sends the matter back unless the RIRs do not accept the case for changes. A second rejection does not simply kill the matter without further institutional consequence.

The Board's no-action route is especially interesting. It can let a policy become effective without an affirmative Board vote. That is efficient where the regional and ASO record is strong. It is also a form of accountability risk. If the Board allows acceptance by inaction, the record should still show that the Board had the file, that the window ran, and that no concern rose to the level of rejection or change request. Silence may be a valid legal outcome. It should not be a public mystery.

The Board check is legitimate because global number policies require action by IANA or an ICANN-related body. ICANN cannot be forced to carry out an external instruction without any corporate review. But that review must remain bounded. The Board should not use its position to rewrite regional number policy or to favor one region's scarcity theory. Its proper questions are whether the process met the ASO MoU standard, whether the policy is within global number-resource scope, whether IANA can execute it, whether significant viewpoints were considered, and whether the outcome is consistent with ICANN's limited role in number resources.

The post-exhaustion IPv4 policy shows a relatively clean Board outcome. The accepted policy gave IANA a limited recovered-pool mechanism. It did not require the Board to settle regional transfer policy, ownership claims or local distribution rules. That narrowness reduced the chance of Board overreach.

Future Board action should be documented in a way that helps separate legal review from policy substitution. If the Board asks questions, the questions should be published with enough detail to show whether it is testing process, scope, implementability or overlooked viewpoints. If it requests changes, it should identify the exact concern. If it rejects, it should meet the explanation duty with specificity. The Board should be a final constitutional check, not an opaque last desk.

Delay rights are spread across the chain

The global policy process contains many lawful delay points. Regional forums can take time to discuss, translate, meet, run final calls and ratify. RIR staff can take time to reconcile common text. The NRO EC can take time to confirm adoption and referral. The ASO AC has up to sixty days after EC advice to pass the proposal, return it for review or request more time. The ICANN Board has sixty days after receipt to act, and its change request can send the proposal back toward the beginning if at least one RIR agrees that changes are needed.

This is not accidental. Delay can be a safeguard. A global policy with weak regional support should not race into IANA execution. A proposal with unresolved text differences should not be rushed. A Board concern about overlooked viewpoints deserves an answer. In a system that allocates globally unique resources, slow legitimacy may be better than fast error.

But delay also has winners. In IPv4 scarcity, waiting is not neutral. Incumbents with stock can survive delay better than entrants without addresses. Regions with greater remaining inventory experience recovered-pool timing differently from regions under acute pressure. A party that benefits from no global rule can use procedural caution as a strategy. A party that needs a rule may accept weak text to end waiting. Every delay point therefore needs a public reason.

The 2011 post-exhaustion case shows both sides. The earlier failures delayed a recovered-pool policy, but those failures also improved the eventual design. The final formula was narrower and more acceptable because prior versions exposed disagreement. The delay served deliberation. That is the kind of delay the system should defend.

Other delay would be harder to justify. If common text sits unresolved because no owner is named, that is not deliberation. If an EC referral waits without a public clock, that is not deliberation. If ASO AC concerns are described only in broad terms, that is not deliberation. If a Board change request identifies no concrete overlooked viewpoint, that is not deliberation. The difference is reasoned traceability.

A mature global policy record should therefore include a delay ledger. It should show each clock, the body responsible, the reason time is being used, and the next decision point. It should distinguish unavoidable meeting cadence from missing information, text reconciliation, translation, legal review, operational readiness and substantive disagreement. This would not force faster policy. It would force honest timing.

The ledger should also survive failure. A failed proposal often teaches more than a successful one, but only if the public can see where it failed. The 2009 and 2010 recovered-space efforts reveal regional text and consensus barriers. Future failures should be easier to read. Was the veto regional, textual, procedural, executive, Board-level or implementation-based? The answer changes the remedy.

Delay is a form of power. The global policy chain should treat it that way.

Responsibility cannot disappear into consensus

Consensus language can obscure responsibility. When a policy succeeds, everyone can claim the legitimacy of the whole chain. When it fails, each body can point to another stage. A region can say the global text was not ready. The EC can say the regions did not produce identical versions. The ASO AC can say significant viewpoints were not adequately considered. The Board can say it needed changes. The result can be accurate and still leave the public unable to assign responsibility.

The solution is not to personalize blame. It is to identify institutional responsibility at each decision point. A regional policy forum is responsible for its own consensus call, objection treatment and adoption text. RIR staff are responsible for documenting common elements accurately. The NRO EC is responsible for referral readiness, executive coordination and any refusal to proceed. The ASO AC is responsible for process review and its decision to pass, return or request more time. The ICANN Board is responsible for any acceptance, rejection, change request or no-action acceptance under its window.

PTI is responsible for execution once policy exists.

This map should appear in every global proposal file. A reader should be able to look at the status page and know who currently holds the file, what the holder can do, what the holder cannot do, what clock applies, and what public record will prove the decision. Without that map, the global process may remain formally accountable but practically hard to question.

The responsibility map also protects institutions from unfair accusation. If a proposal fails because one region rejects it, ICANN should not be blamed for refusing to impose it. If the ASO AC returns a proposal because significant objections were not considered, the RIRs should not say the Council invented a policy disagreement. If the Board rejects by supermajority with a specific process concern, the Board should not be accused of ordinary policy preference without examining the concern. Clear responsibility makes criticism more precise.

It also protects resource holders and operators. They care less about institutional etiquette than about whether IANA will have a rule and when. If recovered IPv4 space is not being distributed, the public should know whether there is no space, no active trigger, no adopted policy, no common regional text or no Board ratification. Each condition has different consequences.

Responsibility becomes even more significant when policy touches market expectations. A returned-pool rule affects how holders, buyers, networks and public agencies imagine the future supply of addresses. A delay in the rule can influence private transactions. An unclear failure can fuel rumor. A clear failure lets actors price uncertainty more honestly.

Consensus should therefore be treated as a result, not a fog. The fact that many bodies must agree should make responsibility more structured, not less visible.

The procedure needs a public evidence file

Every global proposal should carry a public evidence file with five parts. First, it should state the IANA action required and why the question cannot be handled by regional policy alone. Second, it should list every regional version, adoption status, final-call dates, objections and board or community ratification act. Third, it should show the common text and explain any differences from regional drafts. Fourth, it should record NRO EC and ASO AC decisions, including reasons for referral, return or extension. Fifth, it should record ICANN Board questions, actions, no-action dates, concerns and any mediation trigger.

The ASO proposals page already does some of this work. For GPP-IPv4-2011, it lists the proposal, summary, acceptance status, and regional introduction and ratification links. ICANN's background report provides a richer table of regional steps. Those records are valuable because they let readers trace the policy across regions. But the standard should be made more explicit for every future proposal.

The file should also include a short "why this matters" field. For recovered IPv4, the reason was not simply that a policy existed. It was that returned address space after exhaustion needed an IANA allocation mechanism. For ASN blocks, the reason was allocation from IANA to RIRs. For IPv6 blocks, the reason was central allocation criteria. Naming the operational consequence keeps the proposal from drifting into symbolic language.

The evidence file should mark uncertainty. If a regional archive link changes, say so. If a final-call record exists only in a mailing list, identify it. If the same proposal has local identifiers such as prop-097, ARIN-2011-9 or RIPE 2011-01, map them. If adoption dates refer to community consensus in one region and board ratification in another, distinguish those acts. Precision reduces false equivalence.

The file should not become a barrier to participation. It should be maintained by the institutions that already coordinate the proposal, not by a volunteer proposer alone. The proposer has a duty to help communities understand deliberations elsewhere, but the legitimacy of a global policy should not depend on one person's record-keeping capacity. RIRs, the NRO secretariat and the ASO AC should carry the public file once the proposal is recognized as global.

A public evidence file would make the procedure more readable for courts, governments, operators, researchers and members. It would also help future proposers. They could see why prior drafts failed, which objections mattered and which compromises worked. The record would become institutional memory rather than scattered archaeology.

What the post-exhaustion case teaches for the next proposal

The post-exhaustion IPv4 chain teaches five lessons. First, the IANA action must be narrow enough to execute. The accepted policy described a pool, trigger, period, unit and announcement boundary. It did not try to decide every regional use of scarce IPv4 space. That restraint made global agreement possible.

Second, regional disagreement is not a defect when it produces better text. The 2009 and 2010 failures were costly, but they filtered out designs that could not command identical consent. The 2011 version succeeded by avoiding some of the earlier disputes. The system worked because refusal led to revision rather than institutional paralysis.

Third, every refusal point needs a reason. "Version differences" is a real reason, but it should be broken down. "No consensus" is a real result, but it should identify the material objection. "Withdrawn" is a real status, but it should state why the proponent or community moved away. If refusal is legitimate, it can withstand explanation.

Fourth, the Board's role should remain narrow. The Board ratifies or checks a global number-resource policy that has passed the regional chain. It should not become the forum where regional economic preferences are relitigated. The more complete the regional and ASO record, the less temptation there is for Board substitution.

Fifth, the implementation boundary must be clear. PTI and IANA carry out the global rule. RIRs distribute resources under regional rules. A global policy that blurs that line will either overreach or fail. The recovered-pool rule worked by stopping at the central allocation step.

These lessons matter now because future global proposals may be more politically sensitive than the current small list of active policies. Questions around registry continuity, RPKI trust anchor configuration, resource returns, emergency allocation, sanctions constraints or cross-region transfers could test the chain again. Some may be unsuitable for global policy. Others may need exactly this chain because the IANA or ICANN-related action is central. The difference must be argued with evidence, not assumed.

The procedure is not obsolete simply because global policies are rare. Its rarity is part of its function. It should be used only when the central number function truly needs a rule. But when it is used, the public should be able to follow every gate without being a procedural specialist.

The best test is practical. If a network operator asks why a proposed IANA rule did not happen, the answer should fit on one page: which region declined, which text differed, which Council returned it, which executive referral did not occur, which Board concern controlled, or which clock remains open. If the answer requires reading years of scattered minutes and mailing lists, the procedure is formally open but practically opaque.

The procedure works by limiting everyone

The global policy process is often described as slow. That is true but incomplete. Its deeper design is limitation. It limits ICANN by requiring regional agreement before number policy reaches the Board. It limits each RIR by requiring all regions to accept the same text before IANA receives a rule. It limits the NRO EC by tying executive referral to common regional adoption. It limits the ASO AC by making it a process reviewer rather than a policy dictator. It limits the Board by requiring high-threshold rejection, reasons, resubmission rules and mediation after repeated rejection.

Those limits are why the process has legitimacy. No body can plausibly say it owns global number policy alone. The central address pool is global, but authority over the rules that touch it is shared. The sharing is not symbolic. It has teeth: one region can stop the proposal; the ASO AC can send it back; the Board can reject or request changes; no action by the Board can still count as acceptance after the clock. These are real consequences.

The same limits can hide power if the record is thin. A veto is legitimate when it is attached to a reason and a role. It is suspect when it appears as silence, delay or unexplained coordination. A delay is legitimate when it protects deliberation. It is suspect when it protects no one except the actor already benefiting from no rule. A Board concern is legitimate when it identifies overlooked viewpoints or implementability. It is suspect when it becomes a disguised policy preference.

The post-exhaustion IPv4 case remains the best case study because it shows both refusal and success. The first recovered-space proposal could not survive version differences. The second could not survive regional abandonment and withdrawal. The third became policy because it narrowed the bargain, passed every region and gave IANA a bounded action. That is not a story of frictionless consensus. It is a story of institutional gates doing work.

The next global proposal should be judged by the same standard. Do not ask only whether it succeeds. Ask whether every gate explains itself. Ask whether the regional record shows real consideration. Ask whether common text is truly common. Ask whether the EC's executive role is procedural and visible. Ask whether the ASO AC's review is reasoned. Ask whether the Board's action is bounded. Ask whether PTI can execute the result without inventing policy.

When the procedure works, responsibility is distributed but not lost. That is the standard global number policy needs. The Internet number system can tolerate slow consensus better than it can tolerate untraceable authority. The chain is valuable only if each link can be inspected.

Sources