Summary

  • In a no-cache capture at 21:17:38 UTC on 10 September, RIPE NCC's current October 2026 General Meeting page and its archived May 2026 meeting page rendered the same 37 country rows and the same 231 registered votes. The October page says registration is open; the May page describes a meeting completed four months earlier.
  • RIPE NCC's separate final May voting report records 3,421 members registered and 3,049 members casting ballots, while the results announcement also records 3,049 votes cast. Those final records are not contradicted by this finding. What fails is the meeting identity of the unlabelled country table inside the archive.

Analysis

An archive is supposed to stop moving at the boundary that makes it useful. On RIPE NCC's General Meeting pages, that boundary is presently difficult to see.

The current October 2026 meeting page announces a GM running from 28 to 30 October. Its registration notice says registration opened on 12 August, and the page includes a section titled “GM Vote Registrations Per Country”. RIPE NCC explains that the figures are member votes registered by country, based on RegID. When captured without cache on 10 September, the table contained 37 country rows. Germany had 53, Sweden 30, France 16, the Netherlands and the United Kingdom 13 each, and the United States five. The rows summed to 231.

Those figures should not be read as a verdict on turnout or geographic representation. They are an early registration snapshot, not a final electorate, and the unit is votes based on RegID rather than people, networks, address holdings or national populations. The October table may have been entirely accurate at the moment observed.

The anomaly appears when the current page is compared with RIPE NCC's page for the May 2026 GM. That page identifies a meeting held from 20 to 22 May and otherwise reads as a completed-event archive. Yet at the same capture time its registration section contained the exact same 37 rows and every same country count as the October page. It too summed to 231.

There is a firmer May record elsewhere. The final voting-report page says 3,421 members registered and 3,049 cast ballots. RIPE NCC's results announcement independently says 3,049 votes were cast. The gap between 231 and 3,421 is not evidence that the May report changed or that the October figure is false. The two numbers describe surfaces at different stages. The important fact is that a table which behaves like current registration evidence appeared inside a page whose event was already complete.

The strongest benign reading is straightforward: both meeting pages may be presenting a current projection rather than storing a meeting-specific snapshot. The public pages do not reveal whether this comes from a shared component, an editorial choice, a cache or another implementation. Assigning a technical cause would exceed the evidence. But readers should not need to reverse-engineer a website to learn what event a governance table belongs to.

The table itself provides no visible meeting identifier, capture timestamp, lifecycle label or displayed total beside the country rows. A reader landing directly on the May page can therefore make a reasonable but wrong attribution: that the table records May registrations. A researcher may quote it, a member may compare it with a later meeting, and an analyst may calculate country shares. Each downstream calculation can be internally correct while its event key is wrong.

This matters because the RIPE NCC describes the General Meeting as the forum where members discuss operations and vote on resolutions and Board elections. Registration evidence is not decorative event copy. It helps outsiders reconstruct the participation surface around decisions. When the provenance of that evidence floats between meetings, the signed result can remain perfectly sound while the public audit trail becomes unreliable.

The remedy is a small data contract. Every published registration table should state the meeting or event ID, the time at which the figures were observed, whether the table is LIVE_REGISTRATION or FINAL, the number of rows, the displayed total and the definition or version of the data. At the end of a meeting, the public page should freeze an immutable final country snapshot and link it to the final voting report. A later meeting must use a different key.

Those labels are an editorial proposal, not a change announced by RIPE NCC. Nor do they require disclosure of member identities. Their purpose is to make the aggregate evidence self-describing. A person should be able to preserve a table and answer three basic questions without consulting site code: which meeting, as of when, and in what state?

A correction record would complete the chain. If the May page is repaired, the record should identify the affected URL, the snapshot that was replaced, the reason and effective time, and the durable final reference. Quietly changing the page may make it look right today, but it cannot tell tomorrow's researcher what was visible on 10 September.

RIPE NCC's final May report remains the appropriate source for May participation. The 231-row sum belongs to the live evidence observed on the October surface at the captured time. The governance failure is the absence of a visible boundary between those states. Archives earn trust not merely by remaining online, but by remaining attached to the event they claim to remember.

Sources