Summary

  • AFRINIC’s 2 November extended delegation file contains 35 October-dated IPv4 allocation or assignment rows for 35 distinct holder IDs: 30 allocated, five assigned, and 28,160 addresses in total.
  • In that same file, 24 of those holder IDs also have an allocated or assigned IPv6 record. Eleven have no such row. A pending policy draft would require an IPv6 request with a new IPv4 request when the applicant has no IPv6 space, then demand deployment evidence for later IPv4 requests.

A policy proposal can make a clean sentence out of a messy transition: if you want more IPv4, bring IPv6 along. AFRINIC’s current draft says exactly that at the first doorway, then adds a credible deployment plan and timed tests for subsequent requests. The proposal is still under discussion. Before its authors get to argue about the percentages, AFRINIC’s own registry file offers a more elementary count: who had an IPv6 resource recorded when the October IPv4 entries appeared?

The 2 November extended delegation file has 35 IPv4 rows dated October 2026 with allocated or assigned status. Thirty are allocations; five are assignments. Their address values total 28,160. Each row carries a distinct opaque holder ID. AFRINIC’s format guide says rows sharing an ID within one file belong to the same resource holder. It also says those IDs may change between file versions. So the comparison stays inside this single snapshot: join each October IPv4 ID to any allocated or assigned IPv6 row in the same file.

The result is 24 with a companion IPv6 record and 11 without one. Among the 24, the earliest IPv6 date precedes the IPv4 date for 12; the other 12 first appear on the same date. The split is sharper if the status fields are kept intact: 22 of 30 allocated IPv4 rows have an IPv6 match, while two of five assigned rows do. The country codes and block sizes can be counted too, but neither is a shortcut to the holder’s deployment plan. The official guide says country is the holder’s legal base, which may differ from the network’s location.

The draft tries to do more than collect matching database rows. It would ask a new IPv4 applicant without IPv6 space to request IPv6 simultaneously and present a credible addressing and deployment plan. For additional IPv4, access networks would face a timetable for the share of customers able to receive globally reachable IPv6 delegations: 25% after 12 months, 50% after 24, 75% after 36 and 95% after 48. Content hosts would be measured against the share of their assigned IPv4 space also covered by reachable IPv6 assignments. An operator doing both jobs would have to clear both tests independently. These are different denominators.

The file’s 24 matching holders clear no such test merely by owning an IPv6 row.

That is where a registry’s easy arithmetic ends. The public file lists resource records and dates; it gives no application packet, customer-delegation count, content-host coverage measurement or reachability result for these 35 holders. Eleven blank joins identify the visible starting gap. The draft remains under discussion, while these October records were made under the existing rules. To judge either proposed test, AFRINIC would need the holder-level requests and deployment measurements that the daily file omits.

The temptation is to let the first count stand in for the second. It would make a tidy dashboard: 24 green marks, 11 amber ones, everyone applauds the coloured cells. AFRINIC’s proposed obligation is much more demanding. It attaches consequences to a deployment claim, not to a decorative IPv6 allocation. If the working group advances the draft, it should publish the measurement method and a dated baseline in the same breath. Otherwise the registry could learn how to count paired resources while having no public way to tell whether the second network is switched on.