Steve Allen’s page uses historical plots to explain the relationship between UTC, Earth rotation, and atomic time. It recalls that a proposer of leap seconds warned about risks to automated systems before the first leap second, and that USNO time systems were designed without them. A counter on the page also notes that the legally correct number of elapsed seconds since 1972 can differ between jurisdictions and IEEE standards.

UTC combines two references. The second is based on cesium measurements, while the civil day follows Earth’s rotation relative to the Sun. If leap seconds remain, one calendar day represents one rotation of Earth. Under the alternative described on the page, a day would count 794 243 384 928 000 hyperfine oscillations of cesium-133. The author says this link between clocks and calendars was chosen in the 1960s as a practical answer to conflicting requirements.

The page warns against extending a time scale into periods before its creation. Historical names can suggest continuity that never existed. Calendar reforms such as the Augustan correction of the Julian calendar, the end of Islamic intercalation, and the historical February 30 show how civil time can change. The warning also covers interpretations of proleptic time scales.

Its ΔT plots have an approximately parabolic shape, consistent with a roughly linear change in length of day. The vertex reflects the conventions established by Simon Newcomb. His US Naval Observatory team reduced more than 60,000 observations of the Moon and planets, with data from other observatories. In 1896, national ephemeris directors meeting in Paris selected Newcomb’s formulas for use from 1901 onward. The ephemeris day, and later the ephemeris second and SI second, consequently matched the average day over observations centered roughly on 1820. This made the slope of ΔT near zero around 1820 and its value near zero around 1901.

Another plot marks the origin of GMT. The Royal Greenwich Observatory opened only in 1676, so a scale counting elapsed time from 1601 cannot describe a time scale in contemporary use. Any extrapolation before 1955 may also need to distinguish historical non-uniform solar time from a modern uniform interpretation.

The relativistic figures magnify TCB and TDB variation by a factor of 1,000. Astronomical scales near the top are tied to planetary physics. Practical scales near the bottom use measurements made on Earth. TCB has the most uniform rate in the comparison, but no clock can operate at that rate. TDB follows the current rate of TT and suits planetary ephemerides, while Earth’s motion through the Sun’s gravitational field produces an annual variation of about 2 milliseconds. The plots use TT as zero, effectively corresponding to ET and providing the uniform scale closest to measurements from surface chronometers. Most applications use TDB instead of TCB, and TT, practically UTC, instead of TCG.

For most of history, civil clocks followed the Sun. The page says navigation drove the creation and distribution of the more uniform scales. Civil authorities have also changed legal time by hours, citing the 2000 Olympics in Australia, alleged energy savings in the United States, and Hugo Chávez’s policy in Venezuela. Computers receiving UTC broadcasts therefore already need a way to find the offset from civil time. Zoneinfo usually represents whole-hour or half-hour offsets, but it can represent seconds as well. The right zoneinfo files include leap seconds, and POSIX already requires second-level offsets.

The required software machinery already exists in user space. RFC 7808 provides one working and deployed scheme, while an IETF draft describes another. POSIX 2008 specifies tzset(), the TZ environment variable, and rationale A.4.15. The proposed design would give a uniformly increasing broadcast scale a new international name, such as TI, and keep UTC as a time zone layered on top. This would leave the kernel with a uniform system clock and place leap-second data in zoneinfo.

The page links this proposal to an ITU-R colloquium held in Torino in 2003. Its recommendation was to assign a new name to a broadcast scale without leap seconds. The page presents this as a way to preserve the connection between UTC and civil calendars while giving machines a predictable time base. It compares leap seconds with daylight-saving changes, since both reset clocks to reflect a decision about legal time. If UTC itself abandons leap seconds, jurisdictions using it as legal time will eventually define a day without reference to Earth’s continuing rotation.

The plots were inspired by work from Seidelmann and Fukushima, who published a 1992 explanation of time scales introduced by the IAU in 1976 and 1991. Their source figure compares time arguments and scales from 1950 to 2020, with periodic TCB and TDB terms enlarged by a factor of 100. TAI is the reference, while ET, TDT, and TT are treated as TAI + 32.184 s.