Yomiuri Shimbun

Of about 25,000 postal savings automatic teller machines, 1,186 malfunctioned Tuesday at 8 a.m., when they were due to open for business, the Posts and Telecommunications Ministry said.

All the troubled ATM services were back online at about 1:10 p.m.

The ministry said the problem was caused by a computer glitch stemming from the fact that 2000 is an irregular leap year, the ministry said.

Feb. 29 does not normally fall in years divisible by 100, unless that year is also divisible by 400, as 2000 is.

The ATMs affected were older models introduced in 1996, whose chip installed in its internal clock is incapable of recognizing a leap year, the ministry said later.

Among about 7,200 of this particular model currently in use nationwide, 1,186 machines malfunctioned because reprogramming of their old chips during a regular inspection that started January was not completed by Tuesday, the ministry said.

As post office employees took care of customers over the counter, the problem did not cause serious inconveniences, the ministry said.

A total of 1,869 machines that issue receipt forms for registered mail, located at large post offices, also malfunctioned. The machines printed the receipt date as March 1, instead of Feb. 29.

The ministry believes that the problem occurred because the machines' calendars were not programmed to register Feb. 29.

The Automated Meteorological Data Acquisition System (AMeDAS) of the Meteorological Agency, which provides meteorological data for locations throughout the country, began malfunctioning at a little past 1 a.m. The system gave erroneous hourly rainfall data at places where no rainfall was observed.

According to the agency, the problem was caused by the leap-year computer glitch. Its officials told data subscribers not to use the data. The system returned to normal about 4 p.m.

The problem occurred at terminals of meteorological data editing and transmission equipment at 43 regional meteorological observation sites in western Japan, including Kyoto, Osaka and Fukuoka.

From midnight to 1 a.m., rainfall figures for Kagoshima and Kumamoto were indicated as 950 millimeters and 951 millimeters--equivalent to the amount for half a year--but no rain was observed during the period in those areas. The software of the troubled terminals had not been made Y2K-compliant.

As a result, when Feb. 28 rolled over, instead of displaying data for Feb. 29, data was displayed for Feb. 1, and abnormally large rainfall figures were indicated, together with incorrect wind velocities, temperature and daylight hours.