Skip to content

Early May 2026: small fixes and a catalogue

The first half of May was mostly small breakages caught and closed the same day, and one piece of housekeeping that mattered more than it looked: every hub router's licence brought current and put on automatic renewal. The website briefly became a three-product catalogue.

Written up afterwards

Written in September 2026 from our session notes and commit history, and dated to the end of the period it covers.

What shipped

13 May: hub licences on autopilot. Several hub routers were running on licences that had lapsed or were about to. That doesn't take a hub down on the day, but it blocks updates and leaves the hub quietly fragile. All of them were brought current, and a renewal check now runs every few hours and warns well before any deadline.

14 May: online status for every region. The "who's online" view had only ever asked the London hub, so members elsewhere never showed as connected. It now asks every active hub and falls back to recent history after a hub restarts.

14 May: a three-product catalogue. The site presented Torus alongside two products in development, Retrocolo (managed hosting) and Sandbox SGI (shell access to real SGI hardware), with a register-interest form for each. The homepage hero was rebuilt on 15 May to show all three over the network map. Two days later we changed course; see late May. Sandbox SGI eventually shipped in September as bTen.

Also: the network echo service was renamed necho. The old name still works.

What broke, and what we fixed

  • A setup script that failed on paste. The generated RouterOS script for self-hosted Torus setups could stop partway through when pasted into a terminal, because of how one command was wrapped. It now pastes cleanly in one go.
  • DNS over Torus. For a while, being on Torus could stop a device resolving nekotopia.io itself: an internal zone claimed authority over names it had no records for. Fixed the same day.
  • A dedicated-IP setup that hung. Turning on nSolo could occasionally stall on a stuck connection to the router, leaving the address reserved and the dashboard toggle stuck off. There's now a hard timeout with a retry, and the dashboard repairs the state itself if it happens again. We wrote this one up as a full incident report.

By the numbers

34 changes to the platform repository between 1 and 15 May.