(This was written in April 1998).
I’d like an opportunity to set the record straight before the villagers with the scythes and pitchforks crash down the castle gate on January 3, 2000 (Jan 2 is a Sunday).
I’ve been programming since 1965, and have had first-hand experience designing and implementing computer systems with two-digit years.
It wasn’t “inertia”, as BusinessWeek said recently, that brought us to the Y2K impasse. A careful analysis of the evolution of technology shows that the situation is not as simple as it seems.
My first job was with an insurance company. We had 40,000 policyholders, and the master file had an equal number of 80-column punched or ‘Hollerith’ cards. It took two hours to process the file, allowing for jams which often resulted in a torn or mangled card (the card had to be re-entered on the keypunch). Two passes a day, plus miscellaneous reports, accounted for a full shift.
It was very important not to spill over to a second card, which would have doubled the processing time and required a second shift. The latter was already used by the programmers. Stripping the century was an obvious solution to reducing storage space. Every column was very precious. At one point, we developed a method to store 240 digits in a single card.
When we later converted to magnetic tape, one would have thought it an opportune time to expand to four-digit years. Not so: whenever you upgrade to a new medium, it is precisely when this technology is at its peak cost. Tape storage was 800 bits/inch. Our computer had 32K of memory, so we could not use large blocks of records. The ¾” inter-record gap was often larger than the intervening data. As a result, tape storage was also at a premium, and not the time to get expansive about four-digit years.
Magnetic disc was the same story. If memory serves me right, a 20MB removable disc pack was $500; the disc drive, as big as a dishwasher, was about $10,000, and that didn’t include the controller. Once again, storage was expensive, and information was squeezed into every available byte. Two-digit years remained the standard.
As the years went by, Parkinson’s law applied to storage: information expands to fill all available space. Who wanted to work on boring projects like converting date fields? Especially when there were no tangible short-term benefits to sell management. There were always sexier projects.
Furthermore, converting files means parallel runs, which means duplicate files, requiring a doubling of disc storage. Since IT budgets are always maxed out, where does the money come from for all this redundant storage? And what do you do with it when the conversion is over?
There was rarely a clear-cut opportunity to accomplish a conversion.