Summary

  • ICP-2 turns recognition into an institutional threshold: a candidate must show regional support, neutral governance, technical capacity and a sustainable operating model before it receives authority over number resources.
  • The reform can make that threshold more accountable, but a rule controlled by incumbents can also make entry more expensive and protect the existing structure. Version 2 exposes both possibilities; the July 2026 Recommended Draft is not yet public or adopted.

The most consequential line in a registry-recognition rule is not always the one that names the decision-maker. It is the line that decides who must produce evidence, how much evidence is enough, and whether a refusal can be tested by someone outside the refusing institution.

ICP-2 has supplied that threshold since 2001. ICANN accepted it as a framework for evaluating new Regional Internet Registries. Its tests are substantial: a candidate region should be large and non-overlapping; regional network operators must show broad support; policy development must be open, transparent and bottom-up; the organisation must be neutral, technically capable, financially independent and able to keep auditable records. These are not decorative conditions. A failed registry can interrupt allocation, registration, reverse DNS and the records on which network operators rely.

The economic effect, however, begins before recognition. A candidate must organise a continental constituency, document support, finance staff and systems, demonstrate continuity and persuade organisations already served by an incumbent to migrate. That burden can screen out an unsafe operator. It can also screen out a credible institutional alternative whose benefits would appear only after recognition. The distinction depends on whether the test measures risk or merely rewards the scale already possessed by incumbents.

The Version 2 draft tries to make the lifecycle more explicit. It reaches beyond admission to continuing obligations, audits, continuity and possible derecognition. For recognition, it would require each RIR to assess the proposal independently and publish its reasons. Its initial rule relies on unanimous RIR support, but it creates an independent review route: if only one RIR remains opposed, that dissent can be disregarded when an outside reviewer finds material factual error or inadequate justification and the candidate satisfies the requirements.

That architecture contains a real discipline. Published reasons make a veto visible. Independent review reduces the value of an unexplained refusal. Continuing duties mean recognition is not a permanent certificate issued once and forgotten. Yet the same architecture preserves a strong incumbent role. Existing RIRs judge the organisation that may redraw service regions, enter their coordinating body and alter the distribution of authority. Even a justified system can carry an incumbency premium.

LACNIC is a useful institutional test not because there is evidence of a pending rival application—there is none in the closed source set—but because its regional legitimacy is expressed through members, open policy participation and local governance while recognition and derecognition remain globally coordinated. Today LACNIC is a membership-based, not-for-profit registry headquartered in Uruguay. Its published policy process is open to anyone and moves through public discussion, forum debate, consensus, last call, Board ratification and appeal.

Regional legitimacy therefore has more than one price. It requires enough local autonomy for members to shape policy in their own institutional and linguistic setting. It also requires global compatibility so addresses and routing records remain coherent beyond the region. A revised ICP-2 should not pretend those aims are identical. The useful question is whether a candidate can meet shared continuity and accountability outcomes without being forced to reproduce every internal method of an existing RIR.

Version 2 gestures toward that separation. It anticipates separate implementation procedures jointly developed by the RIRs and ICANN. It also adds demanding tests: recognition should not harm the registry system and should produce a material improvement over the status quo. These can be prudent safeguards. They can also shift the burden from “show that you can serve the region safely” to “prove, before entry, that you will outperform institutions that already possess scale, data and coordination relationships.” That is a much higher economic bar.

The reform's current status matters. On 28 July 2026, the ASO Address Council delivered a Recommended Draft to the NRO Executive Council. The announcement says the NRO EC will review it and coordinate the path toward a final version. It also says the Recommended Draft, rationale and redline will be published separately. Until those documents appear and adoption is complete, Version 2 is evidence of the design debate, not the operative replacement for ICP-2. Claims about its unanimity rule or its precise safeguards must remain attached to the draft.

The Network Rights doctrine offers a normative check, not evidence: authority over shared numbering infrastructure should be legible, contestable and tied to continuity for resource holders. Applied here, that means publishing criteria, evidence, reasons, review outcomes and implementation costs. It does not prove that any RIR has acted wrongly, and it is not an ICANN or LACNIC policy.

The best recognition rule will neither romanticise entry nor sanctify incumbency. It will demand credible operations, broad regional support and a financed continuity plan. It will also expose conflicts, permit proportionate independent review and distinguish common outcomes from institutional cloning. That is how a standard disciplines registries without turning recognition into a moat.

Sources

Additional fact-packet sources