Search RPD Archives
Limit search to: Subject & Body Subject Author
Sort by:

[rpd] Updated Policy Proposal - IPv6 as a criteria in IPv4 Soft Landing AFPUB-2026-v6-001-DRAFT03

NP Petronella NPertuniaPetronella at outlook.com
Thu Nov 19 17:01:39 UTC 2026


Hi everyone,

I agree with the concerns Owen DeLong and Mike Burns have raised, and I oppose AFPUB-2026-v6-001-DRAFT03.

Jordi is right that some networks still need IPv4 — for NAT64, for dual-stack business customers, for bringing a new network into service. That need is real, and I do not dismiss it. What I disagree with is the remedy: making access to IPv4 depend on an assessment of the applicant's IPv6 deployment. IPv4 requests should be assessed on demonstrated need, under clear, published and predictable rules.

The central problem is the role this gives AFRINIC. A registry keeps records: who holds which number, under which published procedures. It is an address book, and an address book is not an instrument of enforcement. This proposal would let staff decide whether an operator's IPv6 rollout is good enough before issuing an IPv4 allocation. That is a different function: a discretionary gate over an operational decision that belongs to the network that carries the cost and the risk. What needs coordinating here is uniqueness and accurate records, not an operator's IPv6 roadmap. Withholding an administrative service in order to change an operator's behaviour is not a technical necessity; it is enforcement, and a voluntary registry has no authority of that kind. The registry would gain the power to refuse and would bear none of the cost of the delay that follows. Rules that carry consequences should be made and enforced by bodies that have lawful authority to do so.

The thresholds also do not measure what they claim to measure. For access networks, the result depends heavily on customer premises equipment and on customers' own choices; an operator can enable IPv6 and still fall short because subscribers have switched it off. For hosting providers, having IPv6 assignments alongside IPv4 space says little about whether services are actually reachable, or used, over IPv6. Either way, the metric can be satisfied on paper by a network that has not deployed, and failed by a network that has.

The proposal also works against the operators already doing the right thing. A network running NAT64 is deploying IPv6; it would now need a "credible plan" and a retrospective review to obtain the IPv4 it needs to keep that service running. Meanwhile, a convincing plan is not evidence of deployment. The test selects paperwork, not progress, and it adds cost and delay for the operators who have already made the investment.

On brokers, smaller allocations do not touch the underlying incentive: the gap between the administrative cost of registry space and its market value. Smaller blocks do not create more addresses. They create more applications, more queue position and more discretion over who is served, which rewards those best able to work the process and pushes operators with genuine need into transfers. Mike is right that AFRINIC-issued space cannot simply leave the region — outbound inter-regional transfers are not available except for legacy space, or space previously transferred in, and that limited policy has not been implemented — and that addresses should go to those who need them. Owen is right that the remaining pool should be allowed to run out, as pools have elsewhere, rather than being prolonged with further conditions.

IPv4 should not be the instrument for IPV6 deployment, and holding IPv4 hostage to an IPv6 scorecard will not produce deployment. It produces permission, delay and dependence — and the operator, not the registry, pays for the delay. If we want more IPv6, measure it, train for it and reward it, on its own merits and with its own impact assessment. That is a separate piece of work. It should not be the price of an IPv4 allocation.

Oppose as drafted.

Nonhlanhla

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20261119/ea49877a/attachment-0001.html>


More information about the RPD mailing list