Summary
- The rationing rule is printed to the last digit: first-come-first-served, exactly one /24, a lifetime cap of 256 addresses per LIR, a lifetime eligibility bar, contiguous space or no allocation at all, and no guaranteed waiting time.
- The registry prints the queue's length and the age of its head in dated monthly updates — 806 LIRs and 487 days on 23 October 2026 — but no document I read prints how many /24s are issued from recovered space in a period, and the annual allocation totals are registry-wide aggregates that cannot be joined to the queue.
The block that ended and the list that began
The sentence is exact, and it is in the announcement's own first line: "Today, at 15:35 UTC+1 on 25 November 2019, we made our final /22 IPv4 allocation from the last remaining addresses in our available pool. We have now run out of IPv4 addresses." That is how the last-resort supply of one of the world's five regional internet registries ended — not with a price, not with an auction, but with a block size and a clock reading (ncc-announce, 25 November 2019).
The same day's news item said where the next supply would come from — organisations that have gone out of business or closed, and networks returning addresses they no longer need — and set the terms in one sentence: "Only LIRs that have never received an IPv4 allocation from the RIPE NCC (of any size) may request addresses from the waiting list, and they are only eligible to receive a single /24 allocation." (the run-out announcement).
Read that sentence twice, because the word doing the work is "any". A member that took a /22 in 2005, or a /24 from this queue last year, is outside it for good. This is not a supplier of last resort for members who need addresses; it is a one-time grant to members who have never had any, at 256 addresses each.
What the policy actually fixes
Current policy, ripe-826, published 20 June 2024, puts the whole waiting-list rule in section 5.1 and leaves very little to anyone's discretion (ripe-826). Requests go on "a first-come-first-served waiting list" and "No guarantees are given about the waiting time." "The size of the allocation made will be exactly one /24." "The sum of all allocations made to a single LIR by the RIPE NCC is limited to a maximum of 256 IPv4 addresses (a single /24)." And if a contiguous /24 cannot be assembled, nothing is allocated until enough contiguous space has been recovered.
The operational explainer adds three conditions the bare policy text does not carry (how the waiting list works): the request must come from a member with no prior RIPE NCC IPv4 allocation of any size, including one received through this list; the allocation "must be used to make assignments"; and the space must be used on a network with at least one active element in the RIPE NCC service region. Requests are filed in the LIR Portal and join the queue automatically.
So the terms a reader needs are almost complete. Size: fixed at 256 addresses. Frequency: once, ever. Order: arrival. Position: visible to the member in the LIR Portal. Time: unknown, and printed as unknown.
The clock runs on the space, not on the requester
What the policy does not fix is the service rate, and it cannot, because the rate is a property of the returning blocks. Recovered space is held in quarantine for "six months or as long as the space is globally routed" before it is treated as new address space (quarantine for returned resources). Returned blocks of a /24 or larger re-enter the section 5.1 rules; smaller blocks go to the reserved pool for internet exchange points, so they never reach this queue at all. And the registry is blunt about its own expectations of volume: "It is impossible to predict how many addresses we will recover over the coming years – we do not expect that it will be many." (how the waiting list works).
That gives the mechanism its shape. The wait is not administered by anyone; it is produced by a quarantine clock ticking on blocks nobody outside the registry can count, filtered through a contiguity test. "Recovered" and "available to this queue" are two different quantities, and only the first is described in public.
What is printed, and when
The waiting-list page says plainly what it shows: the number of requests already on the list, and the number of days the LIR at the front has been waiting, in a table and graph "updated every three hours" (IPv4 waiting list). On 20 November 2026, in two separate retrieval passes, those counters came back as "Loading..." and no figure. So the numbers below are the dated printings, not a live reading:
- 23 December 2025 — "LIRs on IPv4 Waiting List: 925 Days that first LIR in queue has been waiting: 563" (member update, December 2025).
- 29 July 2026 — 753 LIRs on the list, 454 days for the first LIR (member update, July 2026).
- 23 October 2026 — 806 LIRs, 487 days (member update, October 2026). The same update prints 20,717 LIR accounts, 20,031 members, and 2,173,440 IPv4 addresses transferred in September 2026 — a transfer figure, on a different metric from waiting-list recovery.
The arithmetic: the front of the queue changed hands
Three printings, two subtractions, and both are mine.
Between the 29 July and 23 October 2026 updates, the front LIR's elapsed wait rose from 454 to 487 days: 33 days. The two updates are 86 days apart. So over 86 calendar days the head of the queue aged 33 — and the queue itself grew by 53, from 753 to 806.
Go back further and the direction flips. Between 23 December 2025 and 23 October 2026 the head aged from 563 days down to 487: a fall of 76 days across 304 days. The elapsed age of one continuous request cannot fall. Therefore the request at the front of the list in October 2026 was not the request at the front in December 2025. The head changed at least once inside those 304 days.
What the published documents do not say is why. An allocation, a withdrawal, an entry that was administratively reset — any of them would do it, and the registry prints no count of any of them. What can be said is narrower and still useful: the position is not a stable clock, the list's length is a net figure in which arrivals and departures are merged, and a fall in the count is not a count of /24s issued. 925 to 753 is minus 172; 753 to 806 is plus 53. Neither is a service rate.
The annual totals that will not join to the queue
The RIPE NCC Annual Report 2025, published as ripe-855 on 17 April 2026, prints a two-year table: 723 /24 IPv4 allocations and 185,088 IPv4 addresses allocated in 2025, against 732 /24 and 187,392 in 2024 (ripe-855). I read those figures from a search excerpt of the PDF rather than in a PDF viewer.
The arithmetic closes both ways, and that is the interesting part. 723 × 256 = 185,088. 732 × 256 = 187,392. Each address total is exactly the /24 count multiplied by 256, and divides back exactly. So the annual aggregate carries no unit other than the /24 — and it still cannot be joined to the queue, because the report does not say how many of those allocations came from recovered space and how many were transfers, leases or internal re-allocations. A reader holding a queue position in the hundreds cannot divide 723 by anything to reach a date.
Two locks on the same block
The block a member receives is not freely disposable. Resource transfer policy ripe-807, published 30 October 2023, holds that scarce resources "cannot be transferred for 24 months from the date the resource was received by the resource holder", and the restriction travels with a change of business, including a merger or acquisition (ripe-807). The membership page puts waiting-list /24s under that clock explicitly, and carries a dated restriction of its own: "As of 29 September 2026, the RIPE NCC has temporarily restricted existing members from acquiring additional LIR accounts", until further notice, with members who already hold several accounts not required to close them (become a member). The page does not state a reason for the restriction and I do not supply one.
Put the two side by side and the entry route narrows without any price moving. A /24 from this queue is a two-year-locked 256 addresses, one per member, available only to members with no allocation history — and a member cannot simply add a second account to try again.
What a reader can compute
Almost everything, which is why the gap is worth naming precisely. Entitlement is printed. Block size is printed. Order is printed. Position is visible. The live table claims a three-hour refresh, and the monthly updates put two integers on the record every month. What is never printed is throughput: how many /24s the registry issued from recovered space in a given period, and how much space entered and left quarantine.
One addition would cost nothing and change the whole picture: the entry date of the request at the front, rather than only the days it has waited. A composition change would then be visible as a change of date rather than inferred from an arithmetic reversal. Add reissued /24s per period and quarantine inflow and outflow, and a member could recompute an outcome instead of waiting to be told one. Until then this is a queue whose length and head age anyone can read, and whose speed only the registry can see.
Boundaries
- The reading is bounded to the twelve documents cited here, retrieved on 20 November 2026 between 13:29Z and 13:41Z. Every absence claim is scoped to that set and date.
- No live queue value is quoted. The live page's counters returned no numbers in any pass on 20 November 2026; only dated member updates carry values, each labelled with its publication date.
- The monthly printings are dated snapshots. No document read defines a reporting period for them beyond the publication date, and none states whether the counted list holds the same population from month to month.
- The 723 /24 and 732 /24 annual totals are registry-wide aggregates, not waiting-list counts, and no document read reconciles them to the queue. They are not used here as throughput.
- The head-of-queue turnover is my arithmetic on three printed integers. Its cause is not established and I allege none.
- Changes in queue length are net flows. Neither a rise nor a fall is presented as the number of allocations made.
- The annual report figures came from a search excerpt of the PDF and were not read in a PDF viewer.
- The operational explainer carries no printed publication date and can be updated independently of the policy document; where the two differ in detail, both are reported and neither is adopted as the other's correction.
- Returned blocks smaller than a /24 go to the reserved pool for internet exchange points, so recovered space and queue supply are different quantities.
- Nothing here concerns any named member's request, position, allocation or conduct. The registry publishes no queue identities and none is inferred.
- The 29 September 2026 account restriction is reported as printed and dated, beside the one-/24 rule. No cause, motive or connection is alleged.
- No BGP session, looking glass, routing-table query, RIPE Database query or traffic measurement was performed. Nothing here states where any address is routed or used.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






