The discipline of the timetable
Northline Rail's 1974 national timetable was more than a schedule — it was a masterclass in constraint-solving that software still gets wrong.
The 1974 Northline Rail passenger timetable was not designed by software. It was assembled by a team of twelve scheduling clerks in a windowless room above Crewe station, working with printed train graphs, wax pencils, and a rule that no two trains could occupy the same section of track within a four-minute headway. The result was a document spanning 1,350 pages, coordinating 18,000 daily services across 2,500 stations, with a measured punctuality of 92 per cent. No algorithm has since matched its reliability at that scale.
The timetable as a constraint graph
The Crewe re-signalling project of 1985 demonstrated what happens when local optimisation is mistaken for system design. Engineers replaced the mechanical lever frames with computer-based route-setting, reducing signal-box staffing by forty per cent. But the new system treated each junction as an independent decision point, ignoring the inter-train dependencies the timetable had encoded by hand. Within six months, late-running services doubled.
Every cancelled service was a constraint the timetable had already absorbed.
What the Crewe clerks practised — and what the re-signalling engineers discarded — was a form of constraint propagation that remains poorly understood outside railway operations. A delay at Doncaster was not a local problem. It was a signal that rippled through seventeen subsequent services, reaching Edinburgh by midday and Penzance by evening.