Time is an illusion. Lunchtime, doubly so.

A Civil Time Alternative to the Leap Hour

Sophie’s view, in summary: Keep UTC continuous and correct daylight through the tried & tested time zone machinery that we already have.

The decision to make UTC continuous is, in essence, sensible. Treating an eventual one-hour divergence from the turning Earth as a future leap hour is not. We already have a mature, widely deployed mechanism for translating reference time into human time: time zones.

Proposal: Correct time at the layer that owns the requirement

Some images in this article
are generated by

EU AI Marker icon

EU AI Act Regulation 2024/1689

The conclusion first

The requirement to keep civil noon somewhere near the middle of the sunlit day belongs to civil time. It does not belong to an atomic reference scale. That sounds like a fussy distinction until someone proposes solving a daylight problem by making the world’s reference clock jump by an hour.

My preferred architecture is simple. TAI continues to count atomic seconds. UTC remains contiguous. UT1 continues to describe the rotation of the Earth. When the accumulated difference becomes socially unacceptable, every civil time zone changes its UTC offset by one hour in the corrective direction. Civil time is corrected; the underlying clocks remain unchanged.

Key argument. Most general-purpose operating systems, mainstream programming platforms, phones, and networked services/devices already support a time zone setting and receive time zone rule updates. This mechanism is not perfect, but it is real, deployed, routinely exercised, occasionally exorcised, and designed for precisely the job in question: turning a continuous global instant into the local clock time chosen by a jurisdiction.

A leap hour would be an enormous new failure mode invented, it would appear, to rudely avoid using an existing configuration mechanism.

**What has actually been decided**

In October 2026, the General Conference on Weights and Measures adopted Resolution 5:

Continuous UTC is to take effect on 20 May 2027, and the permitted magnitude of the difference between UT1 and UTC will become 3,600 seconds. The resolution explicitly recognises that critical infrastructure needs continuous UTC, that many civil clocks already differ from apparent solar time by more than an hour, and that daylight-saving arrangements regularly change clocks by an hour.^[1]

The 3,600-second figure is a tolerance, not a detailed instruction to insert or delete 3,600 seconds when the boundary is eventually approached. Recent reporting has, somewhat understandably, labelled the destination a ‘leap hour’, but that is more of a memorable description of the unresolved future problem than it is a settled implementation design.^[2]

That distinction really matters. We have time to decide what should happen at the boundary. We should not allow a catchy phrase to congeal, like a half-eaten breakfast, into an architecture.

Four layers with four different jobs

Timekeeping becomes much less mystical once the layers are allowed to perform their own jobs.

Layer	Purpose	Treatment
TAI	Uniform atomic time	Continue without civil adjustment
UTC	International reference time and the basis for most civil time	Keep continuous from 2027-05-20, as Resolution 5 intends
UT1	Observed rotation of the Earth	Measure and publish the difference from UTC
Civil time	The local clock used by people, law, transport, and commerce	Change time zone offsets when daylight alignment requires it

IERS already publishes UT1 minus UTC as an Earth-orientation parameter, including regular estimates and predictions. Specialist users do not need a damaged UTC in order to know where the Earth is pointing.^[3]

Move the offsets not the countries

Suppose that, after centuries of accumulated drift, UTC is one hour ahead of UT1. At an internationally agreed instant, each civil zone would reduce its offset from UTC by one hour.

Thus Europe/London would remain Europe/London, but its standard-time rule would change from UTC+0 to UTC-1. Europe/Paris would change from UTC+1 to UTC+0. Europe/Bucharest would change from UTC+2 to UTC+1. Seasonal daylight-saving offsets would move by the same amount.

The local clock would repeat one hour at the transition. Relative differences between those regions would remain unchanged. UTC timestamps, elapsed durations, logs, satellite systems, and atomic time would continue without a jump. Historical local times would remain historical facts because time zone databases retain their previous dated rules.

If the accumulated drift were in the opposite direction, the signs would reverse and the local clock would skip an hour. This is why the proposal should be specified as a rule change tied to the observed sign of UT1 minus UTC, not as having a pre-determined direction.

Do not rename the zones. Do not move their geographic boundaries. Do not rewrite history. Keep the stable identifiers and add a new rule from a defined UTC instant. Do not pass GO – do not collect £200.

The machinery already exists

The IANA Time Zone Database records the history of local time and is periodically updated when governments alter UTC offsets, boundaries, or daylight-saving rules. The data within is machine-readable, and operating-system vendors distribute it to users.^[4]

Windows maintains time zone definitions and updates them through the operating system. Android uses IANA data and ICU translations. Java’s ZoneId is built around named regional zones such as Europe/Paris, with rules supplied by the IANA database. Apple’s Foundation framework likewise represents a time zone as a geopolitical region with an offset from Greenwich Mean Time.^[5][6][7][8]

Thus the essential abstraction is already on billions of devices. A properly engineered system stores an instant in UTC, stores a named time zone where local meaning matters, and calculates the displayed wall time from the rules applicable on that date.

This is the crucial difference between Europe/London and UTC+00:00. The first is an identity whose rules can change over time. The second is merely a fixed offset. A future event saved with a named zone can follow a lawful rule change. A future event saved only as a naked local time, or with a hard-coded offset, has thrown away the information needed to behave correctly.

Those weak systems already fail when governments change daylight-saving dates or national offsets, as the 2007 extension of US daylight saving time demonstrated.

They are not evidence that the time zone mechanism is unsuitable. They are evidence that some software has been written as though politicians, geography, and daylight would remain conveniently static.

This is not free. But it is bounded

A coordinated civil-time correction would still be a non-trivial global change. There would be one repeated or missing local hour. Timetables, future bookings, payroll cut-offs, recurring meetings, batch schedules, cron jobs, market sessions, and legal deadlines would need explicit semantics. Old embedded devices would need attention. The date line and unusual extreme offsets would deserve particularly robust testing.

But this is a category of change that we already understand. Daylight-saving transitions create gaps and folds every year. Governments change time zone rules. Vendors ship databases. Calendar systems distinguish between an instant and a recurring local appointment. Android asks governments to provide at least six months’ notice even for ordinary rule changes. A once-in-several-centuries correction has no excuse for arriving as an emergency. The policy and technical architecture can be agreed far in advance. The precise transition instant should be fixed as early as Earth-rotation predictions permit, with ample implementation notice (years rather than anything smaller).^[5]

The comparison is not ‘disruption versus no disruption’. It is a bounded discontinuity in civil presentation versus a discontinuity in the reference scale consumed by infrastructure that has no interest in sunrise, lunchtime, or any concept of bedtime.

A civil correction asks systems that display local time to perform local-time work, whereas a literal one-hour adjustment to UTC asks every system that consumes UTC to participate, including systems that should never have been in the daylight business at all.

How I would specify the correction

The policy would need international agreement because civil time is controlled by governments. That is a governance challenge, not a technical objection. A leap hour would also require international agreement, only with a much wider technical (and indeed potentially literal) blast radius.

1. Define the trigger. Publish the relevant UT1 minus UTC threshold, the direction of correction, and the body responsible for recommending action. Keep the scientific measurement and the civil decision distinct.

2. Choose one global instant. Specify the transition by an unambiguous UTC timestamp. Every participating zone changes its offset at that same instant, preserving the relative differences between regions. Europe already coordinates seasonal changes at a single instant, 01:00 UTC. Before the spring step, that is displayed locally as 01:00 in London, 02:00 in Paris and 03:00 in Bucharest.

3. Preserve zone identities. Keep identifiers such as Europe/London and Europe/Bucharest. Add a future rule to each zone rather than forcing users and applications to select a different zone name.

4. Give very long notice. Announce the policy years ahead, publish provisional data early, and freeze the final rules on a timetable agreed with IANA and major mobile-platform vendors, airlines, finance, telecommunications, and critical infrastructure.

5. Publish conformance material. Provide machine-readable rules, test vectors for instants around the transition, examples for every offset, and clearly defined treatment of gaps, folds, recurring events, deadlines, and interval calculations.

6. Audit the bad patterns. Find fixed offsets masquerading as zones, local timestamps stored without provenance, stale appliances, handwritten conversion tables, and scheduled jobs that assume every civil day contains exactly 24 distinct wall-clock hours. These are already broken, so fix them.

7. Leave the past alone. Time zone databases are historical records. The correction applies from its effective instant. Earlier timestamps must continue to render using the rules that actually applied when they occurred.

Reasonable objections

Is this simply global daylight saving? Mechanically, it uses the same kind of offset transition. Institutionally, it would be a one-off civil epoch correction with international coordination, not a seasonal policy.
What about astronomy and navigation? They can continue to use UT1 and published Earth-orientation data. The BIPM resolution explicitly points to the availability of UT1 minus UTC through IERS, GNSS messages, and metrology services.^[1][3]
What about countries already far from solar time? A uniform correction preserves their existing political choice relative to neighbouring zones while removing the accumulated global drift. It does not pretend that every parliament currently optimises for local noon.
What about future appointments? Systems must know whether an event is anchored to an instant or to local wall time. That distinction already exists for flights, broadcasts, recurring meetings, and daylight-saving changes. The correction makes it important; it does not invent it.
What if a government refuses? Then its civil clock diverges from the coordinated scheme. That risk is political and must be managed by treaty, standards, and notice. It is not cured by corrupting UTC for everybody.

Verdict

Making UTC continuous is the right decision. It ends the repeated operational hazard of leap-second adjustments. The one-hour tolerance places a generous bound on the resulting divergence from the turning Earth.

The 2012 issues were instructive; the requirement was published but too many production systems nevertheless treated the edge case as somebody else’s problem.

The unresolved question is what civil society should do when the permitted divergence eventually becomes noticeable.

The answer should not be to save up a century-scale software incident and hand it to our descendants with a cheerful label. We have spent decades teaching computers that local time is a political overlay on a continuous instant. It would be perverse to protect that overlay by damaging the reference beneath it.

UTC continuous. TAI continuous. UT1 measured and published. Civil time corrected through the time zone rules already used to express civil choices.

Keep the clocks that measure time.

Adjust the rules that tell humans when lunch happens.