Nobody Owns the Timekeeping

Almost everything scheduled, logged, billed, or transmitted depends on a shared answer to a question that has no natural answer: what time is it, exactly, everywhere.

That answer is assembled, not measured. Dozens of laboratories run atomic clocks, report their readings, and a weighted average is computed and published after the fact. No clock is correct. The average is correct by definition, and every participating laboratory then steers itself toward it.

There is no authority that compels participation. There is no penalty for leaving. The arrangement is, in the most literal sense, voluntary — and it has held for over half a century.

The budget problem is not theoretical

Veyra spoke with staff at six national timing laboratories across four continents. All six described the same trajectory: stable or declining real budgets, an aging cohort of specialists, and increasing difficulty explaining the work to funders who assume, reasonably, that time is already solved.

One laboratory has a single remaining engineer qualified to service its primary standard. Two have deferred instrument replacement past the manufacturer’s recommended interval. None described the situation as a crisis, and all six used some version of the phrase “it is fine for now.”

Why it keeps working

The honest explanation offered most often was professional pride, which is an uncomfortable thing to find load-bearing in critical infrastructure. The people who do this work believe it matters, know that almost nobody else does, and have arranged their careers around that asymmetry.

The second explanation is structural, and more reassuring. Because the system is an average, it degrades gracefully. A laboratory that drifts is quietly down-weighted. A laboratory that stops reporting simply stops contributing. There is no single point whose failure propagates, which is why the arrangement has survived funding cuts, institutional reorganizations, and at least one national timing service being shut down and restarted six years later.

1. 01Participation is voluntary, so nobody can be forced out for political reasons.
2. 02Contributions are weighted, so a degraded clock harms the average very little.
3. 03Results are published retrospectively, so there is no real-time dependency to attack.
4. 04The output is a correction table, not a service, so it cannot be switched off.

It is a design that would almost certainly not be approved today. It has no service-level agreement, no single accountable operator, and no revenue. It also has an uptime record that commercial infrastructure rarely approaches.

The engineers we spoke to were consistently more worried about their successors than their equipment. Instruments can be bought. The knowledge of how a particular twenty-year-old standard behaves in July, when the building’s air handling struggles, cannot.