Summary

  • ARIN-2026-2 would replace the current LIR follow-on tests with a 75% threshold measured against prefix units in the applicant’s two-year plan.
  • The draft asks applicants to state typical prefix sizes but never says how a /48, /56 and /64 count toward that threshold.

Analysis

The number 75 survives. The thing being measured does not.

Under the current NRPM, an LIR can qualify for a subsequent IPv6 allocation through any of three routes: use at least 75% of its total address space; use more than 90% at any serving site; or allocate more than 90% of its total space to serving sites, with each site’s block justified under the initial-allocation rules. ARIN’s Draft Policy 2026-2 replaces that menu with a planning test. Its current text is marked Under Discussion, not adopted policy.

The proposed test starts with a network plan. Applicants would describe typical assignment sizes—ARIN gives /48, /56 and /64 as examples—set out the network hierarchy and project utilization around years one, two and five. For an additional allocation, the LIR would need to show that it has assigned or reallocated at least 75% of the “prefix units” in its current two-year plan. It would also document deployment against the hierarchy it submitted, explain any architectural change, and provide a revised plan covering a new five-year horizon. ARIN says it would extend an existing allocation contiguously whenever possible.

That is a meaningful change in the ruler. Today’s 75% test looks across total address space; the draft’s 75% test looks at a forecast the applicant itself defined two years earlier. A plan can reflect the shape of a network better than a single pool-wide percentage. It can also turn a forecast into an audit trail: what the applicant said it would build, what it actually deployed, and how much of the planned capacity it has consumed.

Then comes the loose screw. The draft asks for typical prefix sizes, but it does not define how differently sized prefixes become “units” for the 75% calculation. A /48 contains 256 /56s, and a /56 contains 256 /64s. Are those each one unit because each is one assignment, or does the calculation normalize address capacity? The text supplies no conversion rule. That distinction could change when two applicants with the same number of customer assignments reach the threshold.

This is not a complaint that planning is too technical. It is a request for the arithmetic to match the ambition. ARIN’s own planning guide favors hierarchy, aggregation and growth projections; the draft takes those ideas seriously. But a threshold is only portable between applicants when they count the same thing. “Seventy-five percent of the plan” is crisp until one plan counts customers and another counts differently sized prefixes.

The draft also links the next block to proof that the earlier hierarchy was implemented, or a technical justification for changing it. That gives the review a second job alongside utilization: assess whether the network followed its own blueprint. The revised five-year horizon then asks the applicant to explain what comes after the current two-year meter runs down. Planning becomes both the sizing case and the evidence file for the next request.

ARIN discussed the draft at its October 22 public-policy session; the current proposal page still lists it as Under Discussion. The question for the next revision is straightforward: keep the plan-based model, but specify the unit. Otherwise, the clean new percentage inherits an old policy habit—making applicants guess which denominator the policy has in mind.

Sources