Continuous Timescale to Spur Telecom, Internet Efficiencies Remains Elusive

GENEVA -- Talks on introducing a continuous timescale to stem the costly and error-prone process of inserting leap seconds are widening beyond the technical community in the hope of finding an acceptable solution before the 2015 World Radiocommunication Conference (WRC-15), which will decide on the matter, participants said during a two-day workshop held by the ITU and the International Bureau of Weights and Measures. The frequency and complexity of inserting leap seconds is expected to rise in the future, participants said. Currently a leap second is inserted about once every 18 months. The potential leapsecond pitfalls also appear bigger as communications and other systems grow in complexity, they said.

Timekeeping and the implementation of leap seconds affect fixed and wireless telecom networks, navigation systems, satellite networks, radiocommunication services, aeronautical and safety services, financial services, Internet routing and applications, speakers said. The leap second was introduced in 1972 because it was deemed a big improvement over previous methods of coordinating atomic time with the Earth’s rotation, speakers said. Adjusting to the introduction of leap seconds is costly and prone to error for users who depend on accurate time, speakers said.

It’s “absolutely not obvious” what the solution to the problem will be at WRC-15, said Vadim Nozdrin, an ITU-R counselor for science services, during a news conference. Not many administrations have been involved in ITU-R talks on the matter, Vincent Meens, chairman of the ITU-R study group on space science services told us.

The U.S. has supported abolishing leap seconds. The U.K. has opposed eliminating them. Russia has consistently opposed suppressing the use of leap seconds, said Meens, chief of the bureau of frequencies at Centre National d'Etudes Spatiales in France. Russia was very difficult in ITU-R meetings this month, said an official who was not authorized to speak to the press on the matter. He referred to Russia trying to shape draft text in the WRC-15 preparatory report to support its position. What the U.K. eventually decides will be a very important element in the process, said an official who wasn’t authorized to speak to the press about the matter.

The international atomic timescale, Coordinated Universal Time (UTC), as defined by the ITU, is used worldwide, said Francois Rancy, director of the Radiocommunicaiton Bureau. It’s disseminated by radiocommunication systems, he said. An ITU-R recommendation on standard frequency on time frequency emissions is incorporated by reference into the Radio Regulations, he said. It provides the “official regulatory, legal definition of the time,” he said.

The current definition calls for UTC to be a stepped atomic time-scale, said Ron Beard, chairman of the ITU-R working party that deals with terrestrial and satellite time and frequency signal services, and which is drafting approaches to address the WRC-15 agenda item. It’s adjusted with leap seconds to maintain it in close agreement with the Earth’s rotation, said Beard, who works at the Naval Center for Space Technology.

Leap seconds cause widespread use of nonstandard timescales, said Judah Levine, a physicist at the National Institute of Standards and Technology Time and Frequency Division. The financial industry, in arbitrage trading for instance, depends on rapid transactions, said Levine. Time intervals are measured in milliseconds, he said. This becomes “a significant issue,” he said. NIST doesn’t have an official position on discontinuing leap seconds, he said.

Computers deal with adding a leap second by repeating the last second of the day, which creates ambiguities, Levine said. The roughly 75,000 requests per second for time in standardized format received by NIST servers deliver the same results over a two-second period, he said. Digital time stamps thus can reverse causality, he said. Inserting the leap second on the subsequent day results in adding the leap second on the wrong day, which creates other problems, he said. About 50 percent of commercial network time protocol servers work this way, he said. Amortizing the leap second over time does not relate to standard times, he said referring to Google’s approach.

Amortizing the leap second introduces both a “time error and a frequency error” during the time when the amortizing is being done, he said. Google is “very proud” of its approach, he said. “If you use Google time and expect that you're traceable to national time standards, the answer is ‘you're not,'” he said. “You'll discover that when you wind up in federal court,” he said. “That happens,” he said. “Google has not done it right,” he said.

Federal court cases have already dealt with these types of questions, Levine said. The cases involve insider trading, which is fundamentally a time-based crime, he said. “It’s an action which is legal in some times, and illegal at other times,” he said. “When someone buys a security on insider trading, the question of when they did it is of primary importance,” he said. “They are either guilty or innocent depending on the accuracy of that time stamp and the traceability to national standards,” he said.

Major computer operating systems either don’t immediately recognize leap seconds or they apply them incorrectly, Levine said. The next network update for a computer could be a week later, he said.

Some argue that the present definition of UTC should be kept so applications critical to human life can operate as they do now, Rancy said. Another perspective is that the UTC timescale without leap seconds will diverge from the Earth’s rotation time, he said. This would fundamentally change society’s understanding of time and may result in technical difficulties for some applications, he said.

Several arguments have been made for the adoption of a continuous reference timescale that would abolish the leap second, Rancy said. One of the arguments is that the insertion of leap seconds is a costly process that reduces the availability of systems that depend on time, he said. Stopping all clocks in the world for one second creates an ambiguous hiatus where orderly processes such as precise time stamping are disrupted, he said. The difficulty and frequency of inserting leap seconds will mount in the future, speakers said.

ITU-R studies are examining the possibility of achieving a continuous reference timescale for radiocommunication systems, Rancy said referring to consideration of the matter at WRC-15. WRC-15 will consider the feasibility of achieving a continuous reference timescale, either by modification of UTC or some other method, Rancy said. The question of removing leap seconds has been discussed in ITU-R and other organizations for more than 10 years, speakers said.

A single common reference time would be an advantage to those who need it, Beard said. Alternatives that have been discussed include: 1) create a new timescale, possibly called International Time to replace other timescales; 2) use International Atomic Time (TAI), but that’s a metrological scale and isn’t distributed; 3) use GPS time as the official international timescale, but there are problems with the idea of using time associated with systems; 4) introduce leap minutes instead of seconds after a bigger time interval, or; 5) stop the insertion of leap seconds. The U.S. supported removing leap seconds in the run-up to the 2012 Radiocommunication Assembly, where the matter was discussed, speakers said. Consensus there still hasn’t been reached, speakers said.

The U.K. developed a policy at a high government level to support retaining leap seconds and reiterated that policy over successive administrations, said Peter Whibberley, a senior research scientist at the National Physical Laboratory in the U.K. Ending leap seconds is perhaps the most significant change in centuries for civil timekeeping, he said. It would permanently break the link between civil timekeeping and the Earth’s rotation, he said.

There is a lack of evidence of “severe” problems with the insertion of leap seconds, Whibberley said. The U.K. government said technical solutions should first be explored, he said. The U.K. government said there’s been no study of how the divergence would be corrected, he said.

There would be an increasing divergence between UTC and the Earth’s rotation over time, Whibberley said. One hour in difference could add up in about 900 years, he said. One option for addressing that would be to use leap minutes or leap hours, he said. Bigger steps would be costly to implement, he said. A second option is to change time zones once 60 minutes of divergence between UTC with no leap seconds and the Earth’s rotation had accumulated, he said.

The U.K. government is more concerned with the short-term rather than long-term effect of abolishing leap seconds, Whibberley said. The definition of a day would be changed in the mind of the public, he said. It would be fundamental change of what time is all about, he said. It may lead to social, legal, or religious effects that haven’t been explored, he said. There’s already a fairly high level of suspicion in the U.K. of science and technology “being imposed” on the public, he said.

Other less drastic options are available, Whibberley said, referring to using both TAI and UTC. GPS time or other system time could also be used, he said. A continuous timescale would thus be available for applications that need it, he said, but UTC with leap seconds would be available for civil timekeeping. Another approach could be to announce leap seconds further in advance, he said. That would make software changes easier, he said. The U.K. in 2014 will hold a public consultation to broaden debate outside the technical community and internationally, he said.