Summary
- The IETF's own meeting-network page publishes exactly four prefixes and credits the sponsors who donate each meeting's circuits. On the pages reviewed for this article, it prints no autonomous system number at all.
- The number exists anyway. An IETF-published NOC deck from 2012 records the meeting networks "advertised via AS# 56554", and RIPE NCC's own RIPEstat renders the holder of that AS number as "IETF-MEETING Internet Society".
- The durable asset is documentary, not physical. The assignment and the AS number survive the dismantling of the network in every city; the transit is the perishable part, and a different carrier gives it away almost every time.
I commissioned this piece to test a claim that turned out to be half wrong, which is usually where the interesting half is. The question was whether the published record shows the IETF meeting network as an independent numbered operator — its own address block, its own autonomous system, its own upstream — rather than a VLAN riding the venue's ISP. Three of those four hold. The upstream does not, and never did.
The page that lists everything except its own name
The current IETF Meeting Network Information page is a working document, not a statement of identity. It says the dedicated NOC team works with venues before and during each meeting, and it lists the space:
- 2001:67c:370::/48
- 2001:67c:1230::/46
- 31.133.128.0/18
- 31.130.224.0/20
Four prefixes. Two in IPv6, two in IPv4, and a sentence that follows them: "The IETF NOC supplies connectivity for the conference space and the venue hotel." The same page, in the version retrieved for this article, credits Comcast, SFMIX, Hurricane Electric and CoreSite as connectivity sponsors for IETF 127, with Juniper Networks and Cisco as contributors. It carries no publication date. It is updated per meeting, which means it is a current-state snapshot that will read differently after the next one.
That is the whole of the public operational disclosure a reader gets from the front door: four prefixes, a list of donors, and no routing identity. The Meeting Technology page is even more compressed. "IETF meeting networks have redundant, multi-gigabit connections provided by sponsors for each meeting location," it says, and then sends the reader to the NOC site for detail.
This is not a concealment. It is a division of labour, and it is worth naming precisely because it is easy to mistake for an absence. The publisher tells you which numbers it uses and who paid for the pipe. It does not tell you, on those pages, which autonomous system announces them.
What the 2012 deck printed, and what changed since
Go back to the IETF 85 NOC report, delivered in Atlanta in November 2012, and the convention is the opposite. The slide deck pairs the carrier with the number: "2x 1 Gb/s link to Comcast" on one line, "Networks 130.129.0.0/16 and 2001:df8::/32 advertised via AS# 56554" on the next. The IETF 84 report from Vancouver does the same thing with a different carrier — "2x 1 Gb/s link to Telus" — and the same AS number.
The 2012 deck also contains the line that makes this article possible: "Advertising new official address space: 31.130.224.0/20, 31.133.128.0/18, 2001:67c:1230::/46, 2001:67c:370::/48." Those are, to the octet and the nibble, the four prefixes printed on the current page. The two blocks announced alongside them in 2012 — 130.129.0.0/16 and 2001:df8::/32 — are gone from the current list. So the change across thirteen years is a subtraction of borrowed transit space, not a change in the IETF's own four blocks.
The pages reviewed do not document every year between those endpoints; they do document both ends, and the four prefixes match.
Trace the circuit donors and a pattern emerges that no single page states. Comcast at IETF 85. Telus at IETF 84. The IETF 125 Shenzhen page, dated 30 October 2025, reads: "The IETF 125 network connects to the Internet with 2x10Gb/s donated by China Mobile and 2x1GB/s donated by China Unicom for Transit." Comcast, SFMIX, Hurricane Electric and CoreSite for IETF 127. The carrier changes because the venue changes. The address does not, because it is not the venue's to give.
The registry, and the awkward fact that it is behind glass
Here the reporting gets less tidy, and the reader deserves to see why.
The registry of record is the RIPE Database, because the address space is RIPE NCC's. IANA's IPv4 address space registry lists 031/8 as ALLOCATED to RIPE NCC in 2010-05, and its IPv6 unicast assignments table lists 2001:600::/23 as ALLOCATED to RIPE NCC on 1999-07-01. The IETF's space sits inside both. For IPv4 that is a direct line in the table. For IPv6 it is arithmetic on my part: IANA's table is organised in /23 blocks and carries no separate row for 2001:67c::/32, so the containment is a calculation, not a quoted IANA statement about that prefix. I would rather write that sentence than let a tidy chart imply an authority that did not say it.
The RIPE Database web query interface returned no readable text to the tools used for this review — it is a JavaScript application, and neither the AS56554 object nor the prefix objects could be quoted from it. What I have instead is RIPE NCC's own RIPEstat page for AS56554, which renders the name and holder as "IETF-MEETING Internet Society", and two third-party reproductions of whois output. The first, a mirrored aut-num record, shows AS56554 with as-name IETF-MEETING, description IETF Meeting Network, organisation ORG-IS136-RIPE, status ASSIGNED, created 16 March 2011, and a sponsoring-org reference. The organisation object on the same page reads org-name Internet Society, country CH, registered number CHE-100.233.697. The second, a mirrored inetnum for 31.133.128.0/18, shows netname ietf-meeting-network, status ASSIGNED PI, created 10 May 2011 — and country AT, which contradicts the CH on the organisation object.
Two mirrors, one disagreement, and one detail that matters more than the country field: the status word. The AS number is ASSIGNED. The address blocks are ASSIGNED PI. They are different objects carrying different statuses, and any article that merges them into "the IETF's provider-independent AS number" has stopped reading.
What "provider independent" actually promises
The current RIPE IPv4 policy, ripe-826, published 20 June 2024, defines the label in one sentence: "ASSIGNED PI: This address space has been assigned to an End User for a specific purpose. It cannot be used to make further assignments to other parties." Read it twice and notice what is missing. The word provider does not appear. Neither does independent. The policy constrains purpose and forbids further assignment; it says nothing about who may carry the traffic or whether the holder must own a building.
What the policy does not do is sever the assignment from its criteria. The same document ties continuing validity to the original criteria remaining valid and the assignment staying properly registered — a reading of its section 6.3, reported rather than quoted, because that is the honest description of a provision I read in a retrieved rendering rather than a page image. The older, now-obsoleted ripe-622 put it in words closer to the marketing: the space "can be kept as long as the criteria for the original assignment are met".
That phrasing is exactly what people mean when they say provider independent, and it is not the current text.
So the status the IETF's blocks carry is narrower than the story usually told about it, and more useful. It is a purpose-bound assignment to an end user, portable across carriers, not resold or sub-assigned. That is a fair description of a network that changes city three times a year.
Where the permanence actually lives
Put the documents side by side and the mechanism is plain. The IETF requirements draft, published 21 May 2025, intended status Informational, author J. Daley of the IETF Administration LLC, and expired on 22 November 2025, describes the network's equipment and, in the middle of a list, this: "Dedicated IPv4 and IPv6 address blocks that are ‘owned’ by the IETF." The quotes around owned are the draft's own typography, and they are doing real work. The draft will tell you the blocks are dedicated and, in scare quotes, owned. It will not tell you a prefix or a number.
The IETF blog post announcing the new NOC lead, published 10 August 2026, describes the operator in organisational terms — volunteers from the IETF community, a specialist contractor called Linespeed, the remote participation provider Meetecho — and records Joe Clarke's appointment in 2026. Not one address.
Lay those against the registry and the architecture of the thing becomes legible. The organisation behind the number is Internet Society, a Swiss-registered entity. The network is run by a volunteer corps and two contractors. The circuit is given away by a local carrier for a week. Only the numbering is continuous, and the numbering is the one part nobody at the venue controls.
This is the answer to the question I started with, and it is sharper than the one I expected. The IETF does not demonstrate that the address system can provision temporary operators. It demonstrates something better: that a temporary operator can hold permanent numbers, because permanence in this system is a documentary status and not a property of the wire. The prefixes were new official space in 2012 and they are the space today. Somebody registered them once, to an organisation that still exists, with a purpose that has not changed, and the fact that the equipment gets packed into flight cases every few months is beside the point.
That is worth knowing the next time someone argues that only permanent infrastructure deserves permanent numbering. The IETF has been running the counterexample at three meetings a year, in three countries, without ever printing the number it depends on where a reader would look first.
The number is not hidden. It is filed.
Evidence note
Primary pages from the Internet Engineering Task Force and its NOC: the meeting-network page, the technology page, the IETF 85 NOC deck, the IETF 84 NOC report, the requirements draft, the IETF 125 meeting page and the 2026 NOC lead post. Registry and policy material from RIPE NCC and IANA: ripe-826, RIPEstat AS56554, the IANA IPv4 registry and IANA IPv6 assignments. Where the registry itself would not open, the reproduction used is named: mirrored aut-num and mirrored inetnum. The entity is recorded in the BTW directory as Internet Society, IETF Meeting Network.
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.






