Sunday, November 6, 2016

Daylight Savings Time

At the end of the 1980s, I worked on a project that managed a number of Data General MV/Eclipse minicomputers. The operating system that ran on them had the peculiarity of not tolerating the system time being set back: setting the time back would crash the process that managed queues and logins. Therefore on the autumn Sunday when the country reverted to standard time, operators would change the time and restart the computers.

Eventually, Data General modified the executive program to survive the time being set back. Eventually, too, it was possible to connect to a government service over the internet and set the time automatically. (This was in the days before one heard much about the Network Time Protocol; and I'm not sure that the DG minis ever supported NTP.) The time changes became more routine, less a source of chaos and confusion. Computer clocks have become more accurate, and the use of NTP more widespread, so that now one expects the systems to handle time and time changes correctly and without fuss. However, when the rules for daylight savings time change, one must patch operating systems and other programs that use local time.

The United States Congress has an odd belief that extending daylight savings time benefits the economy. This belief is not, as far as I know, shared by statisticians and economists. During the oil shock of 1973 and 1974, the country stayed on daylight savings time for the winter. I was commuting to college then, leaving the house many days by 6:30, and got in a fair bit of star-gazing those mornings. More recently, in 2005, daylight savings time was extended to the first weekend of November. My chief reaction then was annoyance at having to apply patches for no good reason.

Today's New York Times carries an opinion piece by James Gleick, a historian of science, suggesting that we should all be on Greenwich time, or UTC as some call it. That is fine advice for computers: run the system clock on UTC, and display local time according to the location of the machine or (when it can be inferred) the user. For the rest of us it seems too radical.

No comments:

Post a Comment