Summary

  • RIPE proposal 2024-01 would allow expected utilisation over the next 24 months to justify an IPv6 PI assignment larger than a /48. That forecast is an admission input, not yet a complete compliance cycle.
  • The RIPE NCC’s impact analysis says monitoring would require significant administrative work and that the consequence of partial or incomplete implementation after 24 months is unclear. It identifies a choice between potentially disproportionate deregistration and tolerance for an undefined period, without defining a middle decision path.

A two-year plan creates a date. It does not, by itself, create a decision.

The draft text for proposal 2024-01 says that a request for an IPv6 Provider Independent assignment larger than a /48, an additional assignment or an extension must justify the need. Expected utilisation within the next 24 months is one of the examples it gives. The rule would make a forward-looking plan part of the evidence used at the request stage.

The proposal is not current policy. The proposal page and the current-proposals register list it in Review Phase. The policy-announce notice asks for comments before 23 September 2026 and says the working-group chairs will assess rough consensus after the review. The RIPE NCC Executive Board strongly advises revision, but that advice is neither adoption nor a veto.

The strongest criticism comes from the RIPE NCC’s own impact analysis. It translates the utilisation language into a 24-month implementation plan for needs beyond a total /48. Registration Services would have to evaluate those plans. Later, the RIPE NCC and the resource holder would face significant administrative work to verify or monitor whether End Sites and routing policy were implemented as documented.

Then the analysis reaches the missing state: the consequence of partial or incomplete implementation after 24 months is unclear. Its legal section asks the same question more directly. If the expected utilisation does not happen, should the extension or additional assignment be deregistered? The public draft supplies no graded answer.

That absence matters because forecast error is not one thing. A deployment can be late while the underlying need remains real. Demand can shrink. One site can be complete while another is blocked. Evidence can be stale, inaccurate or deliberately false. A holder can cooperate with a review or ignore it. Treating those states as equivalent would convert a planning horizon into a blunt expiry test that the text never actually defines.

The impact analysis describes the institutional bind. Identified non-compliance could lead the RIPE NCC to apply its closure and deregistration procedures, potentially ending in resource deregistration and, in some cases, membership termination. It warns that this may be disproportionate to the violation. Tolerating non-compliance for an undefined period, however, would weaken the credibility of the IPv6 policy and its enforcement.

The current closure document, RIPE-858, is a serious endpoint. It contains notice, response, evidence, arbitration and deregistration paths for established grounds. But the existence of that procedure does not answer the proposal-specific question: what degree and cause of variance from a 24-month utilisation plan should enter that path, after what opportunity to cure, and on whose recorded judgment?

RIPE-738 remains the current IPv6 policy baseline. Proposal 2024-01 would replace it only if accepted. Nothing in the public record establishes that a holder has already failed this proposed test, that the 24-month point is an automatic expiry, or that deregistration would be inevitable.

The missing artefact is a post-plan decision record. A privacy-safe receipt could bind the request, policy version and a digest of the plan to its milestones, observation date, realised need, material variance, cure period, response, deciding authority and review path. It would not need to publish the applicant’s network design. It would need to show which rule transformed evidence into an outcome.

Without that record, the policy has a clock but no state machine. The applicant knows what it must forecast. The registry knows it may have to monitor. Neither can point to a published rule for moving from “plan partly realised” to “update the evidence”, “allow a cure”, “reduce or return space”, “escalate”, or “close the case”.

Sources