On distributed systems, public transport, and the strange comfort of a timetable that's almost right

The London Underground is a real-time distributed system that has been running, with hot deploys, since 1863. It has consensus problems, partial outages, retries, back-pressure, and a dashboard that lives on the wall of every station.

What software people call resilience, transport planners just call Tuesday. A meditation on why the better metaphor for distributed systems isn’t the data centre, it’s the timetable.