Summary

  • In one no-cache capture session, AFRINIC's English current-proposals index returned HTTP 200 and listed seven 2026 proposals, while the official Arabic current-proposals URL returned HTTP 404.
  • The English index labelled three proposals Last call; their detail pages and three RPD notices corroborate the status, identifiers and a closing time of 31 July 2026 at 23:59 UTC.
  • Nothing in the capture invalidates those proposals or their Last Calls. It shows a narrower governance problem: a consequential language route has no visible state, timestamp or explicit path to the authoritative record.

Two routes, two different kinds of evidence

The useful comparison is almost austere. At 06:47 in Shanghai on 11 September, the English page at AFRINIC's current-proposals address returned HTTP 200. Its table contained seven rows for 2026. The first three were Soft Landing Recovered Space and Priority, Amendment of Utilisation in Soft Landing, and a requirement for newly created AS-SETs to use hierarchical names. All three were marked Last call.

The Arabic current-proposals URL was requested in the same session with caching disabled. It returned HTTP 404. It did not redirect to the English index. It did not say that the Arabic edition had been retired. It did not identify a replacement route or display a timestamp for the last available Arabic record.

That is the finding. It is not evidence that every Arabic page on AFRINIC's site is unavailable. It does not locate a faulty application, content-management system or deployment. Without a known last successful response, it cannot establish the duration of the failure. A 404 is a server response, not a causal diagnosis.

The asymmetry matters because the English side is not an empty landing page. It is a compact register of procedural state. Beyond the three Last Call rows, it lists four proposals as under discussion: PDP Working Group Guidelines and Procedures, IPv6 as a criterion in IPv4 Soft Landing, an AFRINIC Policy Compliance Dashboard, and a Dynamic IPv4 Pools Exhaustion Management Framework. The page joins titles, draft numbers, identifiers, dates, states and versions. A reader can use it as an index into the policy record.

The strongest defence is on the English record

There is an important benign account. English is the authoritative working language for much of AFRINIC's policy process. The three proposal pages were available when checked, and each said Status: Last call. The RPD archive was also available. Separate notices from the PDWG Chairs say the proposals entered Last Call after rough consensus at the AFRINIC-37 Public Policy Meeting on 24 June 2026 and give 31 July at 23:59 UTC as the close.

Those records sharply limit the claim. The exact identifiers are AFPUB-2026-IPv4-001-DRAFT02, AFPUB-2026-IPv4-002-DRAFT02 and AFPUB-2026-ASN-001-DRAFT02. The Arabic 404 did not erase those proposal texts, the mailing-list notices or the English index. It does not show that a participant missed the deadline, that a chair acted on the wrong draft, or that any later ratification would be defective. The English index's continuing Last call label after the stated closing date also should not be converted into a verdict about the proposals' ultimate disposition. It is the state the public page displayed at the captured moment.

Nor is a multilingual site obliged to pretend that every version has equal legal force. One authoritative text can reduce ambiguity. Best-effort translations can broaden access without displacing that source. But this defence makes provenance more important, not less. If one language is authoritative and another is explanatory, the explanatory route should tell readers which record controls, how current the rendering is, and where to go when it is unavailable.

A dead route leaves an undated afterimage

The 404 is only half of the evidence problem. Search infrastructure can preserve the other half.

A targeted search during commissioning still exposed an extract from an earlier crawl of the official Arabic URL. The extract showed six proposal rows. Its first three were labelled قيد المناقشة, or under discussion, and its first row carried an older draft identifier than the current English index. That extract is not a fresh page response and is not treated here as an authoritative archive. Its value is narrower: it shows what an ordinary discovery system may continue to present after the origin route stops resolving.

This is how availability becomes provenance. A reader sees an Arabic table but cannot establish its as of time from the live page because the live page is gone. Another reader sees the English table with seven rows and three Last Call labels. Both may believe they have consulted AFRINIC. Only one has a presently verifiable source route.

The usual web convention offers too few states. HTTP 200 means a response arrived; HTTP 404 means the requested resource did not. Governance records need a richer vocabulary. Was the Arabic table temporarily unavailable? Was it deliberately retired? Did its canonical path move? Was generation suspended? Is the last translation still valid as historical evidence? Which English source version did it represent? None of those questions is answered by a generic missing-page response.

The cost is not simply inconvenience. Policy participation is timed. A draft identifier distinguishes one normative text from another. Under Discussion and Last call tell participants different things about the remaining opportunity to frame objections. Even where English controls, an explanatory language surface can shape when a reader realises that a procedural threshold has been crossed. A page need not be legally authoritative to be operationally consequential.

Publish the state of the publication

AFRINIC does not need to promise instantaneous translation to close this gap. It needs to publish the state of the language surface itself.

A small locale-state manifest could sit behind every consequential policy index. It would name the canonical route and locale; the source-record identifier; the source version or proposal-state fingerprint; the time the page was generated; the last successful public validation; the current HTTP state; and the last known successful snapshot. It would also name the authoritative fallback, with its language and timestamp.

The fingerprint matters more than a vague “updated” label. For the current proposals page, it could bind the ordered set of proposal identifiers, draft numbers and public states. A change from six rows to seven, from DRAFT01 to DRAFT02, or from discussion to Last Call would produce a new version. The Arabic edition could then say precisely which English state it reflects without implying equal legal authority.

Failure behaviour should be explicit. If a locale cannot be generated, the route should return a status page in that locale, identify the authoritative English index, disclose the last successful version and make clear whether the older rendering is preserved only as an archive. If the route is retired or moved, a durable redirect or retirement receipt should record the effective time and successor. A search engine can then stop treating an undated fragment as current institutional speech.

The same record would make correction reversible. A repair receipt could name the affected route, the failed or superseded fingerprint, the restored version, the reason category and the effective time. It would not need to expose internal infrastructure. It would simply let readers distinguish absence, delay, replacement and correction.

This is a modest standard for a registry that already asks its community to distinguish policy drafts with exact identifiers. The public interface should exercise the same precision about its own versions.

Sources