RIPE NCC

The entity sits on the left and its connections fan out by role on the right; the graph shows each group’s most confident connections, and the full list below carries every relationship.

Full list

• RIPE NCC Closed the Test API Incident. The Recovery Join Is Missing

  Four status updates show a pilot service moving from investigation to resolution in 37 minutes. They do not show the small set of facts that would let a user connect that green status to a particular test operation, a checked isolation boundary and a completed reconciliation.

  Primary articlePublished 2026-09-20
• RIPE NCC Kept the Lameness Checks and Stopped the Emails

  In 2009, RIPE NCC separated two activities that are often mistaken for one: it continued periodic checks of broken DNS delegations and ended the bulk email alerts generated from them. The decision is a useful historical operating case in what an infrastructure measurement can prove—and what it cannot prove about delivery, operator attention, repair and user impact.

  Primary articlePublished 2026-09-20
• RIPE’s RDAP Marks a Mixed AS Block Reserved at Either Boundary

  A new unit test makes an old classification rule unusually easy to see: a candidate AS-BLOCK needs only to overlap a configured reserved interval for the whole object to receive the `reserved` status. That is a cautious registry rule, but a poor substitute for the status of every ASN inside the block.

  Primary articlePublished 2026-09-15
• RIPEstat Explains 12859:4000 With 12859:4xxx. The Wildcard Is Evidence

  RIPEstat’s Looking Glass can place a useful sentence beside an opaque BGP community. In one frozen response, however, the observed value `12859:4000` received the description “customer routes” from the broader expression `12859:4xxx`. The explanation may be right. The missing distinction is how it became applicable: an observed route attribute, a matched rule and the rule’s provenance are three different facts.

  Primary articlePublished 2026-09-14
• RIPE NCC Has 110 Sanctions Investigations. Its Table Has No Clock.

  RIPE NCC’s latest sanctions report makes the size of its screening work visible: the cumulative register reached 2,172 alerts, with 110 still under investigation on 2 July 2026. What the table cannot show is equally consequential. A reader can see where cases sit, but not how long the open cases have been sitting there.

  Primary articlePublished 2026-09-14
• RIPE WHOIS Gives Invalid API Keys an Audit Identity

  A failed credential usually looks like absence: no user, no permission, no successful update. In the current RIPE WHOIS source, however, the failure path can carry one durable-looking fact into its audit representation—the API key ID supplied as the Basic-auth username—even when validation ends with `Invalid APIKEY`. That ordering turns a rejection into an attributable event, while leaving the governance of the resulting identifier largely outside the code fragment that proves the mechanism.

  Primary articlePublished 2026-09-14
• RIPE’s Plain-Text History Needs Its Request Context

  Whois 1.124 made historical objects easier to read as plain text. The returned body is still filtered history, while the structured response carries a separate revision label. Saving the text is not the same as saving what identifies it.

  Primary articlePublished 2026-09-14
• Sagan Can Mark a Traceroute Successful Without a Reply From the Destination

  One Python flag tests a non-error reply at the end of the supplied trace. Another tests the destination address. Keeping those answers separate is the difference between reading a measurement and awarding it a meaning it has not earned.

  Primary articlePublished 2026-09-14
• RIPE Atlas Tools’ Ten-Second Timeout Stops Before the Atlas API

  A setting added in June reaches the geocoder but not the reviewed measurement-creation path. A bounded source test shows where the budget is lost, not how long any live job waited.

  Primary articlePublished 2026-09-14
• Starla Uses RIPE Atlas Version Numbers. They Do Not Name Its Executable.

  A Rust implementation can speak the probe’s language without becoming the same measuring instrument. Starla makes the distinction visible in source: its own package version, its compatibility numbers and its registration identity answer different questions.

  Primary articlePublished 2026-09-14
• RIPE NCC Is Running Four Security Programmes at Once. Their Handoffs Need a Public Map

  RIPE NCC’s quarterly plan puts audit work, risk management, application security and monitoring on four adjacent lines, all marked “In progress”. That is useful disclosure, but not yet a view of how evidence moves between them. A small public handoff map would let members distinguish four honest partial claims from one falsely reassuring green light—without exposing a vulnerability, a control configuration or an audit workpaper.

  Primary articlePublished 2026-09-13
• RIPE NCC Can Audit IPv6 Resources. Its Policy Does Not Link the Procedure

  The registry’s authority is not the disputed point. The problem is navigational: two IPv6 clauses tell LIRs what they must produce during an audit, while the document never takes the reader to the procedure that explains why an audit starts, how far it can reach and how an outcome can be challenged.

  Primary articlePublished 2026-09-13
• RIPE NCC Lets Operators Sign ASPAs Without Showing the Providers It Sees

  RIPE NCC has made ASPA signing possible and tells operators to list every real upstream. Its public guide also says the Dashboard offers no BGP-derived hint about who those upstreams might be. The missing interface is not an automatic answer but an evidence receipt that helps a signer find omissions without turning observed paths into invented authority.

  Primary articlePublished 2026-09-13
• RIPE’s NRTMv4 Source Gate Reads a Replica. Its Fallback Does Not Cover the Gate

  A three-file code change gives RIPE NCC’s NRTMv4 service a sensible read path and leaves one unusually precise question behind: when a source is new, what proves the catalogue reached that path before the service begins treating it as authoritative?

  Primary articlePublished 2026-09-13
• RIPE Atlas Put Reconnects and Liveness Through One Event Processor. A Slow Backend Blocked Both.

  A RIPE Atlas software probe could have working Internet access on 8 June and still be shut out of the measurement network after a routine disconnect. The public incident record explains why: reconnect work and controller-liveness evidence passed through the same event processor. When a backend slowed, the evidence used to judge controllers arrived late. The failure was not just a queue. It was a queue with authority.

  Primary articlePublished 2026-09-12
• RIPE NCC’s Four Stale IPv6 Glue Records Are Now Three. Root Repair Was Never a Batch.

  One of four old IPv6 glue mappings identified after RIPE NCC’s AuthDNS renumbering has aligned with the nameserver’s current address. The other three have not. That split result reveals the real unit of root-zone maintenance: one delegation, one authorised decision and one auditable outcome at a time.

  Primary articlePublished 2026-09-12
• RIPE Atlas Firmware 5130 Added an OpenWrt Hardware Architecture. The Rollout Needs More Than a Version Number

  Firmware 5130 gives RIPE Atlas a new `openwrt-hwprobe` architecture for next-generation firmware on existing hardware probes. That is a concrete engineering change, not merely a roadmap line. Yet the number 5130 can identify a release without proving which source, toolchain, signed artifact, hardware cohort and update result produced the device that later supplied a measurement. The missing object is a release chain of custody.

  Primary articlePublished 2026-09-11
• A Space IPv6 Draft Gives Every Planet Up to Five RIR Prefixes. It Never Defines the Region.

  The appealing part of a new space-address proposal is that it does not start a sixth registry from scratch. The harder part is hidden in two words. Applicants would use their “appropriate RIR”, but a moving spacecraft has no obvious place on a map built for terrestrial service regions. Before five registries write five sets of policy, the system needs to say what selects one of them.

  Primary articlePublished 2026-09-11
• RIPE NCC’s AuthDNS Ledger Counts 43 Nodes but Not the Eyeball Coverage They Deliver

  RIPE NCC has made the objective unusually clear: put AuthDNS where poorly served networks can actually reach it. Its public inventory can show that 43 hosted instances are operational. What it cannot yet show is whether each new map pin changed the anycast catchment that matters.

  Primary articlePublished 2026-09-11
• RIPE’s IPv6 PI Draft Reserves Growth Space. Partial Transfers Can Make It Unusable.

  RIPE’s current IPv6 PI proposal tries to make future growth easier by reserving a larger nibble-boundary envelope around a new assignment. Its own impact analysis identifies the catch: if the assigned block can be transferred in pieces, the untouched reservation may no longer be usable for the extension it was meant to protect.

  Primary articlePublished 2026-09-11
• RIPE NCC's Archived May GM Page Is Showing October's Live Registration Table

  The October 2026 General Meeting page carried a live country-by-country registration count. At the same instant, the archived page for May displayed precisely the same 37 rows and 231 votes, although May's final report records 3,421 registrations and 3,049 ballots cast. The result is not in doubt; the identity of the evidence is.

  Primary articlePublished 2026-09-10
• RIPE NCC's Admin WebUI Fills Automation Gaps. Retirement Needs an Exception Ledger

  The revealing phrase in RIPE NCC’s current registry plan is not “technical debt” but “admin-like capabilities”: the old interface carries operational exceptions that a cleaner replacement could otherwise erase from view.

  Primary articlePublished 2026-09-10
• A Direct Route Fixed a 150ms Detour. It Has No Backup.

  A route from Britain toward Denmark appeared to turn the Atlantic into a roundabout. The first explanation blamed the wrong control point. Operators then found a short European path, traced the detour to export and peering choices, and installed a direct route whose lack of redundancy leaves the original failure mode waiting in reserve.

  Primary articlePublished 2026-09-10
• RIPE NCC’s New IPv6 Exam Lasts 60 Minutes. Its Badge Will Last Three Years

  A short private assessment is being asked to support a long public claim. Before RIPE NCC launches its IPv6 Advanced certificate, it should make the badge’s changing technical scope visible without exposing a single exam question.

  Primary articlePublished 2026-09-10
• RIPE Atlas Counts Probe Farming by a BGP Prefix. The Cohort Can Move Without the Probe

  A probe can stay on the same shelf, under the same account and on the same access line, yet acquire different neighbours for an admission rule. RIPE Atlas limits software probes partly by the IP prefix “as seen in BGP”. That is a defensible way to protect a shared measurement system. It is also a moving classification, and a decision made from it needs a record that can outlive the route state.

  Primary articlePublished 2026-09-10
• RIPE Atlas Can Show the Edge. It Cannot Show Who Chose It

  A path measurement can show where one request landed. It cannot, on its own, reveal the platform rule that sent it there, the traffic that was eligible, the share served locally or what happens when the appliance disappears.

  Primary articlePublished 2026-09-10
• RIPE NCC’s K-root Refresh Needs Three Acceptance Records, Not One Batch Status

  Amsterdam can be refreshed while the three-site programme remains in progress. That is not a contradiction; it is a warning about the unit of evidence. RIPE NCC has named three K-root core sites whose seven-year-old routers and servers were due for replacement. The useful public question is now not whether K-root stayed online, but whether each site crossed its own change boundary with a reviewable result.

  Primary articlePublished 2026-09-09