Summary

  • The 2012 Applicant Guidebook did not expressly prohibit exclusive operation of a generic string. Its draft registry agreement even allowed an operator to request a discretionary Code of Conduct exemption where all names would be held for the operator’s own use. The central defect was therefore a missing decision test, not an obvious application breach.
  • Charleston Road Registry Inc. filed its .SEARCH change request on 6 April 2013, five days before the GAC’s Beijing advice. The dated record supports a conclusion of strategic adaptation to mounting institutional pressure, but not the claim that the Beijing communiqué ordered a change already submitted.
  • The New gTLD Program Committee controlled the first-round outcome because it could determine whether contracting would proceed and which Public Interest Commitments would enter the agreement. Public comments supplied reasons, the GNSO supplied policy history, the GAC supplied advice, staff implemented instructions, and the IANA functions process later checked delegation readiness. Those were different powers.

The original proposal survived in a redline

The most revealing evidence in the .SEARCH file is not a later policy statement. It is the deleted text in the applicant’s own amendment. The formal applicant was Charleston Road Registry Inc., identified in the application archive as the company behind application 1-1141-50966 for SEARCH. In the 6 April 2013 change-request package, a redline preserved the original model while striking it out: .SEARCH would, in the applicant’s words, add most value by remaining completely closed for Google’s sole use. Second-level registrations would be held by Google, not offered to the public, and the registry would seek an exemption from the Registry Operator Code of Conduct.

The replacement text proposed something different but not unrestricted. The namespace would be dedicated to websites offering search functionality. Registration would be available through ICANN-accredited registrars to parties meeting functional and technical conditions. Registrants would have to support a common query interface; the registry proposed monitoring, audits and enforcement; and it contemplated a redirect service at the top-level label. The revised application, later preserved in ICANN’s public application archive, therefore moved from a single-corporate-group model to a restricted sectoral model.

That difference is narrower than the common contrast between “closed” and “open” suggests. Neither version promised an ordinary retail namespace in which any registrant could use any available name for any lawful purpose. The first version made Google and its affiliates the only possible beneficiaries. The second extended eligibility beyond the corporate group while preserving a search-related purpose, qualification rules and registry enforcement. The institutional dispute concerned where the eligibility boundary ended, not whether every restriction had to disappear.

The chronology matters. The Board’s New gTLD Program Committee had opened a public docket on closed generics in February. Charleston Road Registry said in its 6 April letter that it had already acknowledged the controversy in public comments and was submitting changes through ICANN’s application-change process. The Governmental Advisory Committee issued its Beijing communiqué on 11 April. In the company’s subsequent response to the GAC advice, it described .SEARCH as originally exclusive and the April amendment as a move to restricted access.

The public dates rule out one easy story. The GAC’s Beijing advice did not cause the filing in the literal sense of preceding and directing it. The change request was already on record. A more supportable account is that the applicant adapted to a controversy that had become an operational risk before governments formalised their position. That is still an exercise of institutional power. A gatekeeper can change behaviour through the credible prospect of delay without first issuing a final order. But an inference about anticipated pressure should not be rewritten as a documented command.

The later outcome explains the title. Charleston Road Registry ultimately signed a base, non-sponsored .SEARCH Registry Agreement on 9 June 2016. Specification 11 of that agreement prohibited the operator of a generic string from limiting eligibility exclusively to one entity and its affiliates. The IANA root database records .SEARCH with a registration date of 8 June 2017, followed by a delegation report dated 29 June. Google obtained the top-level domain through its registry subsidiary, but the original sole-group eligibility rule did not survive into the enforceable agreement.

What the 2012 rulebook settled—and what it left open

Retrospective accounts often begin with the eventual contract and project it backwards. The 2012 Applicant Guidebook does not support that move. It did not contain a general provision stating that a generic word could not be operated for the exclusive use of one applicant. On 2 February 2013, the New gTLD Program Committee made the absence explicit: the implemented GNSO recommendations contained no specific policy on closed-generic top-level domains.

The Guidebook’s draft agreement went further than silence. Its Code of Conduct specification allowed a registry operator to request an exemption, in ICANN’s reasonable discretion, if all registrations were held for the operator’s own exclusive use, no registration was sold or transferred to an unaffiliated third party, and applying the Code was unnecessary to protect the public interest. That clause did not promise an exemption. It left the judgment to ICANN and retained a public-interest condition. It nevertheless showed that single-user operation was part of the contractual architecture applicants were told to examine.

A Code exemption was not the same thing as approval of a closed generic. The Code addressed duties such as equal treatment and the relationship between a registry and registrars. If the operator and every registrant belonged to one corporate group, some of those duties would operate differently. The exemption clause did not decide whether a globally delegated descriptive term should be assigned on that basis, whether the allocation would affect a related market, or what evidence would demonstrate a public benefit.

The application could point to a route within the draft contract; it could not point to a pre-published methodology for deciding the larger allocation question.

That missing methodology sat uneasily with the programme’s policy history. The GNSO’s 2007 final report recommended a fair, transparent and non-discriminatory process, with criteria fully available before the process began and normally no later selection criteria. It also called for an orderly and predictable introduction, a pre-published process and contractual conditions capable of securing compliance. Those principles did not eliminate all implementation judgment. They did create a reliance interest: applicants were entitled to ask whether a new restriction was an application of existing discretion or a new selection rule introduced after filing.

ICANN had a competing institutional claim. Applications did not themselves create a right to delegation. The Guidebook included public comment, formal objections, GAC advice, evaluation, contracting and pre-delegation checks. The Code exemption was discretionary. A successful applicant still had to execute an agreement before the string could enter the root. ICANN could therefore argue that it retained room to decline or condition a contract when a disclosed model raised a public-interest problem not fully specified in the policy record.

Both positions had textual support. That is precisely why the case is about a policy gap rather than a straightforward violation. The difficult question was procedural: which body could supply the missing test, at what stage, under what authority, with what evidential burden and with what remedy for an applicant that had relied on the published package?

The Guidebook also separated routes that are often merged in later narratives. Public comments could place information before evaluators, but the Guidebook expressly said they were not formal objections. Formal objections required specified grounds, standing, fees and an expert proceeding. GAC advice was a separate public-policy channel addressed to the Board. An applicant could respond to advice, but that response was not an adjudication before a neutral tribunal. Each process could affect the application, yet each had a different decision-maker and a different remedy.

For that reason, the phrase “the community rejected closed generics” is institutionally imprecise. There was no single binding community vote. Commenters could identify harm or reliance. The GNSO Council could report what the earlier policy process had or had not considered and could initiate new policy work. Governments could advise the Board. Objectors with standing could invoke defined dispute grounds. The Board or its delegated committee still had to decide what those inputs meant for contracting in the live round.

February 2013: the policy omission became a live docket

The 2 February 2013 NGPC resolution was unusually candid. It described the community’s working definition of a closed generic as a generic string proposed to be operated exclusively for a participant’s own benefit. It acknowledged the absence of specific GNSO policy. It then opened a 30-day public-comment forum, requested proposed objective classification criteria, invited the GNSO to provide guidance if it wished, and directed staff to review the policy record, the feasibility of objective classification and the public-interest implications of possible approaches.

The 5 February announcement framed two questions that the filing rules had not answered. How should ICANN determine whether an applied-for string was generic? Under what circumstances should a registry operator be permitted to use an open or closed registration policy? Those were not minor technical clarifications. They were the elements of a decision rule: classification, permission and conditions.

Opening the docket increased transparency. Applicants, competitors, civil-society participants and individual users could see that the issue was unsettled and could put arguments into the record. It did not create an accountability remedy. A commenter could not obtain an order. No independent panel was asked to determine whether reliance on the Guidebook outweighed a later public-interest concern. The forum made uncertainty visible; it did not resolve who bore its cost.

The Board committee’s compressed request to the GNSO exposed a second institutional limit. The GNSO is the policy-development channel for generic names, but consensus policy cannot be produced simply because a live programme needs an answer within weeks. In a 7 March 2013 letter, the GNSO Council said the timeframe did not permit formal guidance with appropriate consideration and consultation. It added that closed generics had not been explicitly considered in the original policy-development process and recalled a broader view that ICANN should not restrict TLD use, leaving applicants to propose open or closed, generic or non-generic models.

That letter should not be overread as a consensus policy authorising every exclusive application. The Council expressly declined to provide formal guidance. Its value was narrower and important: it confirmed that the precise category had not been decided, and it preserved the distinction between a policy-development process and an expedited institutional memory exercise. It strengthened the applicant-side argument that a categorical restriction would be new. It did not remove the Board’s obligation to consider GAC advice or the committee’s practical control over contracts.

The February process also changed incentives before the institution had chosen a final rule. An applicant maintaining exclusive access faced an uncertain contracting hold. An applicant amending to a non-exclusive model could reduce that risk while retaining a specialised namespace. Withdrawal would contain some future expense but surrender the string. The programme’s formal rule had not yet crystallised, but the future approval gate was already shaping present choices.

Charleston Road changed before Beijing

Charleston Road Registry chose adaptation. The 6 April correspondence said .SEARCH would be available to firms offering search functionality rather than reserved to Google. It proposed registrar participation, registrant restrictions, a common technical interface, monitoring and enforcement. The later GAC response called this a “restricted access” model: open to ICANN-accredited registrars, with registrations available to parties that met the registry’s guidelines.

This was not compliance with an already binding closed-generic prohibition. No such prohibition appeared in the 2012 Guidebook, and the GAC advice had not yet issued. The amendment was a strategic use of the application-change process under uncertainty. It removed the feature most likely to trigger a contract hold—eligibility ending at Google and its affiliates—while preserving much of the proposed functional identity of the namespace.

The distinction between an amended application and a signed agreement is essential. An application states what the applicant proposes and what evaluators may consider. A change request alters that proposal. Neither document automatically creates a covenant enforceable after delegation. The .SEARCH contract later made the separation express: Specification 11 stated that commitments, statements of intent and business plans from the application were intentionally omitted.

The April proposal therefore helps explain how the application moved outside the exclusive category, but it does not prove that every search-provider qualification, query-interface promise or service plan became a Public Interest Commitment.

The public record also does not identify a single private exchange that caused the amendment. The applicant cited its own 7 March public comments; government representatives were copied on the April letter; and the Board committee’s docket was already open. Those facts support an inference that institutional pressure mattered. They do not reveal the content of every conversation among Google, ICANN staff, Board members and governments. A careful case file should not substitute a plausible backstage narrative for disclosed evidence.

The power relationship was nevertheless asymmetric. ICANN controlled the only programme route to delegation of the requested string. Charleston Road could amend, wait, withdraw or argue that its model fell outside the emerging category. It could not demand root-zone entry merely because it was technically capable. The applicant retained choices, but those choices were structured by a gate it did not control.

Beijing advice changed the Board’s decision environment

On 11 April 2013, the GAC’s Beijing communiqué advised that, for strings representing generic terms, exclusive registry access should serve a public-interest goal. It listed .SEARCH among a non-exhaustive set of strings the GAC considered generic and associated with proposals for exclusive access. The advice was broad enough to reach a category that the original GNSO policy and the Guidebook had not specifically resolved.

The GAC did not sign registry agreements, conduct technical readiness testing or edit the root zone. Its leverage arose from the Bylaws. The Board had to take GAC advice into account. Under the provisions cited by the NGPC in 2013, a decision inconsistent with the advice required notice, reasons and good-faith efforts to reach a mutually acceptable solution; if no solution emerged, the Board had to explain its final departure. Advice therefore was not a legal veto. It raised the procedural and political cost of disagreement.

That distinction matters because it identifies the implementer. A veto would have allowed the GAC itself to dispose of an application. The Beijing language instead supplied a principle—exclusive access to a generic string should serve a public-interest goal—while leaving key questions unanswered. It did not define genericness, allocate the burden of proof, specify the evidence needed to establish a public-interest goal, identify less restrictive alternatives or prescribe contract language. The Board committee still had to decide how to translate advice into a programme action.

Charleston Road’s response demonstrates the applicant’s limited procedural position. It could submit a written answer, point to the earlier amendment and argue that the factual premise no longer applied because the revised model was not exclusive to Google. That was meaningful participation: the authority controlling the contract would receive the applicant’s account. It was not an enforceable remedy. No neutral decision-maker was authorised by that response process to order ICANN to sign a particular agreement.

The advice should also not be called GNSO policy. The GNSO had not adopted a closed-generic rule through its policy-development process. The GAC was acting within its own advisory function. The operative constraint arose only when the NGPC, acting under delegated Board authority, chose to make non-exclusivity a condition of the contract path and to defer applicants that retained exclusive generic plans.

Two comment records, two different institutional purposes

The public-comment history is easily compressed into one undifferentiated “community process,” but the dates show two records. The closed-generics forum opened on 5 February and closed on 7 March 2013. ICANN staff published the report of those comments on 8 July. It recorded arguments for predictable pre-published criteria and applicant latitude, alongside concerns about competition, consumer choice and the public-interest consequences of exclusive control over descriptive terms.

After the Beijing communiqué, ICANN opened a separate forum on 23 April asking how the NGPC should address the GAC’s safeguards advice. That process closed on 4 June. The 25 June NGPC resolution said the committee had considered the safeguards comments, the closed-generics comments and the GAC advice. The July report was a later staff synthesis of the February record, not the origin of the committee’s June authority.

The distinction is substantive. The February forum asked what the rule should be and how a closed generic should be classified. The April-to-June forum asked how the Board committee should respond to governmental advice in an operating programme. The first exposed a policy-design problem. The second informed an implementation decision. Treating them as the same event obscures the shift from deliberating about a missing standard to using contract authority in response to advice.

The comment record contained legitimate claims on both sides. Opponents of exclusive generics argued that descriptive top-level terms could be common semantic resources, that a single firm might use the allocation to disadvantage rivals, and that dependencies would be difficult to unwind after delegation. Applicants and supporters stressed the Guidebook’s reliance structure, the value of experimentation with different registry models, and the risk that mid-process changes would create winners and losers.

Those arguments did not amount to an empirical finding that an exclusive .SEARCH would harm competition. The effect of a top-level string on a related service market would depend on adoption, user substitution, control of registrations, pricing, technical integration and the availability of alternatives. The record justified precaution and institutional attention. It did not prove anticompetitive conduct by Google, and the analysis does not manufacture present-day registration statistics to fill that gap.

Public comment was participation, not control. Commenters could supply reasons and facts. They could not classify the string with legal effect, order an amendment, insert a PIC into the agreement or delegate the TLD. Nor did publication alone create accountability. A transparent archive can reveal every submission while leaving the decision standard uncertain. Accountability requires an authorised decision-maker, an identified source of power, reasons tied to a standard and a remedy capable of changing conduct or outcome.

The contract gate converted concern into an operating rule

The decisive institutional act came on 25 June 2013. The NGPC recorded that it was acting under authority delegated by the Board on 10 April 2012 to exercise Board authority over issues arising from the New gTLD Program. For applicants not seeking exclusive registry access, it adopted a Public Interest Commitments implementation and directed staff to revise the draft registry agreement. For applicants seeking exclusive access to generic strings, it directed that contracting be deferred pending dialogue with the GAC.

This was the point at which concern became leverage. The committee did not need to reject every affected application or alter the root zone. It controlled whether staff could offer a contract and what terms that contract would contain. Without an executed registry agreement, an applicant could not complete the programme and reach delegation. A contracting deferral therefore operated as a practical stop even if other evaluation work continued.

The resolution and its attached PIC implementation supplied definitions that the Guidebook had lacked. A generic string was a word or term describing a general class of goods, services, groups, organisations or things rather than distinguishing a particular brand. Exclusive access meant eligibility limited to one person or entity and its affiliates. Applicants had to state whether they would accept the PIC language precluding exclusive generic operation or intended to preserve exclusivity.

These definitions made the programme administrable, but their timing carried a legitimacy cost. Genericness can depend on language and context. A term may be descriptive in one market and source-identifying in another. Intended use can affect the public-interest analysis. The committee adopted a contract-facing category broad enough to process applications; it did not conduct a full policy-development process or create an applicant-specific adjudication of every contested classification.

The source of the resulting rule was mixed. The normative concern came from GAC advice and public submissions. The reliance constraint came from the GNSO history and the Guidebook’s promise of predictability. The legal and operational force came from Board-delegated programme authority and the contract. Staff administered the chosen language, but the NGPC supplied the instruction. The applicant could decide whether to adapt, but it did so inside a pathway controlled by ICANN.

Contract was an effective instrument because it joined prospective conditions to post-delegation remedies. A general statement that exclusive access should serve the public interest does not identify a regulated actor or a breach. A registry agreement can define eligibility, registrar access, reporting, dispute procedures, cure and termination. The PIC structure turned the principle into an obligation capable of being tested after delegation.

That effectiveness should not conceal the institutional substitution. Contracting can address a live programme problem faster than a full GNSO process and can attach enforceable terms to a particular delegation. Yet when a contract clause supplies the substantive answer to a category the filing rules did not resolve, implementation begins to perform the work of policy. The immediate risk is contained, but the representative policy process is bypassed at the point of greatest leverage.

Formal objections answered a different question

Two organisations—Initiative for a Competitive Online Marketplace, known as ICOMP, and FairSearch.org—filed community objections to .SEARCH. ICANN’s objection determination record shows that Charleston Road Registry prevailed in decisions dated 4 and 5 February 2014. The ICOMP determination and the FairSearch determination found that the objectors had not met the full community-objection standard, including the required likelihood of material detriment.

Those decisions did not adjudicate the original closed-generic model. The panels assessed the changed application and the defined Guidebook tests for a community objection. They did not decide whether the NGPC could condition contracts in response to GAC advice, whether the applicant possessed a reliance entitlement to its original model, or whether Google-only operation served a public-interest goal.

The remedy demonstrates the boundary. Dismissal removed those objections and returned the prevailing applicant’s advance payment of proceeding costs under the applicable rules. It did not order ICANN to restore sole-group eligibility, execute the applicant’s preferred agreement or delegate the string. Review access was real, but the available relief was limited to the procedural ground invoked.

The .SEARCH file therefore contains several distinct decisions that should not be treated as cumulative votes on one question. Public comments identified concerns. Community-objection panels tested standing and material detriment under a formal dispute standard. The GAC advised on public policy. The NGPC determined the contracting route. The contract set the enforceable operating boundary. Success in one channel did not bind the decision-maker in another.

The 2015 choice removed the original first-round option

The NGPC resolved the remaining round-wide issue on 21 June 2015. It asked the GNSO to include exclusive access for generic strings in policy work for subsequent rounds. It then gave the remaining exclusive-generic applicants three paths: change the application to end exclusive operation and sign the current agreement; maintain the exclusive plan and defer the application to a later round under future rules; or withdraw under the Guidebook’s refund schedule.

For the live round, the practical effect was close to a prohibition. No applicant could retain affiliate-only eligibility for a generic string and proceed to the current contract on the original timetable. Institutionally, however, the committee structured the result as an election rather than a declaration that exclusive generics were permanently unlawful. The preference could be maintained, but the current-round contract would be unavailable.

That design was a compromise among competing risks. Immediate approval would maximise applicant reliance but make a potentially difficult allocation costly to reverse. Outright rejection would maximise precaution but sit most sharply against the absence of an advance prohibition. Waiting for a full GNSO process would preserve policy pedigree while imposing open-ended delay on a live application round. The chosen route allowed conversion, deferral or exit.

The remedies remained asymmetric. An applicant could adapt and proceed, wait without certainty about later rules, or withdraw for a scheduled refund. None restored a timely applicant-specific determination under a public-interest test published before filing, because no such test existed. At the same time, the Board committee did not simply confiscate the application fee and terminate every affected application. It offered a managed conversion or exit rather than full reliance protection.

For .SEARCH, the dated record indicates that the critical adaptation had already occurred. Charleston Road’s April 2013 amendment and GAC response said the application was no longer exclusive. The application therefore sat on the non-exclusive side of the contract line by the time the 2015 election was formalised. That conclusion follows from the public documents; it is not evidence of an undisclosed private settlement.

The 2015 action also revealed an institutional contradiction. The committee acknowledged that the question warranted GNSO policy work for future rounds, yet it resolved the present round through Board-controlled programme administration. The result was operationally bounded and enforceable. It was not equivalent to a consensus policy adopted before applications opened.

The 2016 agreement’s exact answer

ICANN’s .SEARCH agreement page records Charleston Road Registry Inc. as operator, 9 June 2016 as the agreement date, and the contract type as base and non-sponsored. “Non-sponsored” is a category of registry agreement. It does not, by itself, mean that every member of the public can register every available name. The executed terms are the relevant source.

Section 2.9 of the signed agreement required domain registrations to pass through ICANN-accredited registrars, subject to specified exceptions, and required non-discriminatory access to registry services for accredited registrars that entered and complied with the registry–registrar agreement. It also allowed the registry to establish non-discriminatory qualification criteria reasonably related to the TLD’s proper functioning. The contract therefore permitted a restricted namespace.

Specification 11 set the line the original application could not cross. The operator had to publish and follow clear registration policies and operate transparently in accordance with general principles of openness and non-discrimination. For a generic string, it could not impose eligibility criteria limiting registration exclusively to one person or entity and that person’s or entity’s affiliates. The contract did not abolish all qualifications; it prohibited the defining feature of the Google-only proposal.

The same specification made its listed commitments enforceable by ICANN and through the Public Interest Commitment Dispute Resolution Process. It required the operator to implement remedies imposed after a PICDRP determination, including, expressly, the possibility of termination under the agreement. This was not merely a public-interest aspiration. It identified the operator, the prohibited conduct, an enforcement process and a potentially severe remedy.

The agreement also did not convert the entire application narrative into contractual commitments. Specification 11 stated that commitments, statements of intent and business plans from the application were intentionally omitted. That drafting choice prevents the April 2013 amendment from being read as though every element of the proposed search-provider model became enforceable. The amendment helped remove .SEARCH from the exclusive category; the agreement selected which obligations would govern after delegation.

The technical provisions make the same point. The amended application proposed a dotless redirect at the TLD apex. The agreement’s DNS service specification said its permitted zone contents did not allow resource records such as apex A, AAAA or MX records that would enable a dotless domain. A different service would require a Registry Services Evaluation Process request. The application described an ambition; the contract controlled the authorised registry service.

The final arrangement is therefore best described as contractually non-exclusive but capable of non-discriminatory restrictions. Accredited registrars received a right to non-discriminatory access to registry services under the agreement. Registrant qualifications could be imposed when reasonably related to proper functioning. Sole-corporate-group eligibility was prohibited. The detailed 2013 business plan was not incorporated wholesale.

That is as far as the dated record safely reaches. Present eligibility rules, launch status, registration volume or actual market use cannot be inferred from the amended application, the word “non-sponsored” or the fact of delegation. Those questions require current registry policies and operational data. The institutional outcome can be established without inventing a current retail market for .SEARCH.

Delegation was downstream of the bargain

The IANA delegation record lists Charleston Road Registry Inc. as the sponsoring organisation, gives 8 June 2017 as the registration date and links a delegation report dated 29 June 2017. The report records that the New gTLD application process was complete, the applicant matched the contracted party, contacts were confirmed, technical conformance was complete and the remaining processing checks were completed.

The IANA functions process did not reopen the merits of exclusive access. Its role was to verify eligibility, identity, contacts, technical configuration and procedural readiness for the root-zone change. That separation keeps the root function from becoming an ad hoc policy tribunal. It also shows why the visible act of delegation was not the source of the non-exclusivity rule. The relevant bargain had already been written into the June 2016 agreement.

The authority chain was sequential rather than collective in the sense of equal control. The GNSO process supplied the programme’s policy baseline. The Guidebook published procedures and draft contract architecture. The applicant proposed and amended. Commenters supplied evidence and arguments. The GAC advised. The NGPC decided the contract path under delegated Board authority. Staff implemented the agreement. The IANA functions process checked readiness and carried the approved request into root-zone administration. Every actor participated; only some possessed a gate capable of stopping progress at a particular stage.

What the record does not prove

The record does not prove that a Google-only .SEARCH would have harmed competition. The concern was credible because “search” describes a class of online functionality rather than Google’s brand. Yet a competition conclusion would require evidence about market definition, user substitution, adoption of the TLD, the use of second-level names, pricing, technical integration and alternatives. Public comments and an application model do not establish those facts.

The record also does not prove that the April amendment was wholly voluntary or directly compelled. It followed the opening of the NGPC docket and preceded the GAC advice. That timing supports an inference of adaptation to expected institutional resistance. It does not disclose every private conversation or identify a formal ultimatum. The most accurate description is bargaining under asymmetric procedural control.

Nor does the contract establish that .SEARCH became unrestricted. It permits non-discriminatory qualification criteria related to proper functioning and leaves the application’s detailed business plan outside the incorporated PICs. Current eligibility would have to be shown through current registry policy. Delegation alone says nothing about registration volume or commercial uptake.

Finally, the current rule cannot be projected backwards. The 2026-round Applicant Guidebook now states that closed-generic applications will not be approved unless and until an approved methodology and criteria exist for evaluating whether a proposal would serve the public interest. Applicants must affirm that they are not applying for, and do not intend to operate, a closed generic. ICANN’s current explanatory page traces that position to the Board’s January 2024 decision.

That later clarity is evidence of institutional learning, not proof that the same express rule governed in 2012. In fact, the requirement for an approved methodology confirms the original difficulty. A public-interest principle without classification criteria, burdens and remedies did not answer the operational question. The 2012 round supplied an answer through the contract gate after applications were visible.

The counterfactual: a rule at the entrance rather than the gate

A better process would have treated exclusive use of a generic string as a declared application type before filing. An applicant seeking affiliate-only registration would identify the beneficiaries, describe the allocation model, disclose related service markets and accept a defined evidential burden. The application system would then flag the proposal for a distinct evaluation rather than discovering the category through public controversy.

Start with an objective but rebuttable genericness assessment. ICANN would identify the relevant language and commercial or social context, consult recognised dictionaries and sector usage, distinguish descriptive meaning from source-identifying brand use, and publish a provisional classification. The applicant and affected parties could submit evidence. A reasoned determination would explain why the string described a general class or why it did not.

The applicant seeking exclusivity should then bear a defined public-interest burden. It would have to identify a benefit that depends on excluding unaffiliated registrants, not merely a benefit to the applicant from controlling the namespace, and explain why the proposed service could not be delivered through a brand TLD, a restricted but non-exclusive registry, ordinary second-level domains or a time-limited pilot.

Competition analysis should be bounded rather than improvised. ICANN should not attempt a full antitrust proceeding outside its expertise and remit. It could require disclosure of vertical relationships, related service markets, likely excluded registrants and alternative channels; identify facial foreclosure risks; and seek input from competent competition authorities where appropriate. The assessment should neither presume that a generic string creates market power nor wait until every harm has materialised.

Proportionality would connect the evidence to the remedy. The decision-maker would compare affiliate-only eligibility with less restrictive tools: functional qualifications, non-discriminatory registrar access, interoperability commitments, limited reserved names, audit rights or a sunset. If those measures could achieve the stated objective, complete exclusivity would not be necessary. Charleston Road’s own amendment demonstrates the relevance of that comparison: it preserved a search-specific namespace while extending eligibility beyond Google.

Any approval would need enforceable commitments incorporating the representations on which the public-interest finding depended. The agreement would specify eligibility, allocation policies, audit and reporting duties, complaint standing, conflict controls and sanctions. It would say which application statements became contractual and which remained descriptive. A public-interest finding without measurable duties would be a promise, not accountability.

Procedure and remedy would complete the methodology. The rules would identify the decision-maker, publish staff or expert reports, provide an applicant response period, define standing for affected challengers, establish deadlines and state the available relief. A review body should be able to remand a classification error, require adequate reasons or order reconsideration before delegation. Applicants should know in advance whether an adverse decision permits conversion, deferral, withdrawal or a specified refund.

Such a process would not eliminate contest. Genericness is contextual; public interest is disputed; competition analysis can be predictive; and sophisticated applicants will design around definitions. Those are reasons to expose assumptions and allocate burdens, not reasons to leave the gate undefined.

Applied to .SEARCH, the methodology might still have rejected Google-only eligibility. The term describes a broad category of functionality, and a less restrictive model was available. It might instead have authorised a time-limited experiment subject to measurable conditions. The counterfactual does not predetermine the merits. Its advantage is procedural: applicants, governments, competitors and the Board would know the test before fees were accepted, and the institution would have to show how the evidence satisfied it.

The actual process achieved a bounded operational result. .SEARCH was contracted and delegated without an affiliate-only eligibility rule. It did not achieve the same quality of advance notice. The standard emerged through post-application comment, governmental advice and Board-controlled contract leverage; the applicant adapted before the final round-wide options were announced; and the agreement supplied enforceability after the gap had already changed behaviour. ICANN solved the immediate allocation problem through the power of the gate rather than a complete rule published at the entrance.