In 1983, I was in my second year in Canadore College’s computer program. One of our profs asked our class an intriguing question: Why were computer rooms usually windowless? As in 1983, this should be in the software development best practices of 2015 (but since it’s now ingrained, why bother?). Here’s the story he told us, as best I can remember it through the usual college party fog:
Early in his career, he was working in a data center that had large windows, providing a view of the adjacent river and the drawbridge that went over it. People enjoyed working in the data center because of the beautiful view.
Like any data center, there were occasional glitches. One fall, the payroll run aborted at 3:15 in the afternoon. This payroll system ran from tape and cards – one tape was the input tape containing the current payroll records, and the cards were punched with this week’s timesheet information. The program computed the payroll, banged out checks on the printer, and wrote the updated payroll records to the output tape on a second tape drive. But on this occasion, the program crashed part way through, indicating it had hit the end of the input tape, even though it wasn’t close to the end. They called in the repair man, expecting it to be an issue with the input tape drive – specifically the End Of Tape sensor. In these tape systems, the end of a tape was marked with a silver reflective spot that an optical sensor would detect. The technician serviced, polished, and gave lots of love the to tape drive, and all was well again. Until the next payroll run the next week, when exactly the same thing happened, although the stopping point was different.
The technician came back, but couldn’t find anything. So the programmers dug in, searching for anything that could cause such a problem – but came up empty. And that’s the last time it happened – for a year. The next fall, again at 3:15 in the afternoon, there were again failures of the payroll run. And again, after two or three weekly failures, the problem went away until the following year.
By the third year, they decided to switch up the processing schedule, and run payroll at 4:15. As in the previous years, the failures occurred at 3:15, but this time it was a different program running. This was the first clue that would help them isolate the problem.
The next week, someone noticed that the failure occurred exactly when the drawbridge over the river was in motion. At first they searched for a power relationship – was the drawbridge putting noise on the electrical grid which in turn was upsetting the tape drive? Nope – otherwise it would happen all summer (the river froze in the winter, so the drawbridge was active only in the summer).
In the end, a simple curtain solved the problem. It was sunlight bouncing off the drawbridge midway through its arc, then through the glass window at just the right angle, into the computer room, and activating the end-of-tape sensor.
Only in the fall, for a period of two to three weeks, was the sun in exactly the right position for all the angles to be just right to trigger the sensor. There are lots of other reasons for glass-less computer rooms (security, air-conditioning, etc.), but this is certainly the most curious I’ve ever heard.
Lesson: Sometimes computer problems have very odd sources.