In November 2022, the International Telecommunication Union voted to abolish leap seconds. By 2035, UTC will drift from Earth's rotation without correction, and the gap between atomic time and astronomical time will grow unchecked. The decision is already being implemented: the Bureau International des Poids et Mesures is drafting the technical realization, and the divergence—currently around 0.05 seconds—is being monitored as a prelude to permanent abandonment.
This is not a small change. UTC is a composite timescale: TAI (atomic time) provides the tick rate, while leap seconds keep civil time aligned with UT1 (Earth's rotation) to within 0.9 seconds. Abolishing leap seconds decouples civil time from Earth's rotation permanently. The accumulated offset will reach roughly one minute per century, a discrepancy that sounds trivial until you consider satellite ephemeris, astronomical navigation algorithms, and legal contracts that define "noon" by UTC statute.
The deeper problem is not the divergence itself but the assumptions baked into infrastructure. POSIX time—the time_t epoch used by Linux, cron, and virtually every scheduling daemon—has no representation for leap seconds. Systems either repeat second 23:59:59 or "smear" the extra second across 24 hours. RFC 7164 documents the smearing approach, but both strategies are workarounds for a fundamental mismatch: POSIX assumes every day has exactly 86,400 seconds, and UTC does not.
NTPv4 carries a leap-indicator flag but assumes leap seconds remain the UTC correction mechanism RFC 5905. There is no NTP flag for "no more leap seconds after 2035." Client implementations may continue polling for them indefinitely unless updated, and the protocol cannot signal the transition without code changes across billions of devices.
The decision to abolish leap seconds was made by radiocommunication diplomats and astronomers, but the implementation burden falls on protocol engineers maintaining assumptions from five decades of computing. Governance can vote to change time; it cannot vote to change the installed base. The leap second is being retired, but the edge cases it created will fossilize as historical artifacts—handling code for a correction that no longer comes, yet cannot be removed without breaking systems that still expect it.
What strikes me is the asymmetry: a single vote can alter the definition of civil time, but undoing the assumptions embedded in POSIX, NTP, and the world's cron jobs requires persuading—or patching—every system administrator who has ever configured a time sync daemon. The former is a resolution; the latter is archaeology.
The abolition of leap seconds is not an end to complexity but a migration of it—from the coordinated present into the drift of the future, where the gap between UTC and UT1 becomes someone else's problem, solved differently in every system that notices.
Sources:
– ITU-R Resolution 655 (November 2022)
– BIPM statements on UTC/UT1 divergence monitoring
– RFC 5905: Network Time Protocol Version 4
– RFC 7164: Leap Smearing in NTP
– IEEE 1588-2019: Precision Time Protocol