The Millennium Bug – from trilobites to terabytes
Nigel Williams, 10 October 2013
Years before I joined Civitas, I worked as a computer programmer. It was a fragmented industry. My platform of choice was midrange computers built by IBM: the sort of shared, largish computer on which businesses ran their main office functions. In the years before 2000, I and many colleagues spent a lot of time addressing the ‘millennium bug’, with such success that many people still think it never existed.
Computer storage is now mind-bogglingly cheap. People can buy a terabyte of storage (c. 1,000,000,000,000 bytes) just for saving television programmes until a more convenient time, so we easily forget how dear it was before that. In the early 1980s, the mighty University of London Computer Centre charged by the megabyte (c. 1,000,0000 bytes), while home hobbyists had to pay extra if they wanted their ZX81 to have more than 1 kilobyte (1,000 bytes) of memory. It is a commonplace that today’s phones have more computing power than the Apollo 11 mission.
Given the costs, for business systems designed before the 1990s it appeared an extravagance to devote storage to noting the century. The pace of change seemed so rapid that a system would likely be replaced before any other century could happen. The Book of Common Prayer, with its emphasis on ‘world without end’ provided for calculating Easter to 8199 and even beyond, but technology was too locked in an expensive present to look those few decades into the future. As the date grew nearer, systems lasted longer than first planned and it became a possibility that some would read a year ‘00’ as earlier than a year ‘99’.
Obviously, some of the problems got talked up but many were real. Where I was working, the company intellectual, the late Dr Keith Reid, devised some ‘meta-software’. Its input was computer code and its output a list of places where century information was lost or ignored. Our team applied this software to customers’ systems and weighed up how much the faults would matter. One food grower’s system was so neatly written, the work of a single expert, that the only slight weakness was where it had to match an external banking interface. A rental company was in worse condition, being in line to stop sending any invoices. We uncovered such an extensive catalogue of faults for a lender that they abandoned their mothballed system for new applications. Other companies decided to wait and see. At one client, management ceased to receive a report because the system thought it had provided a more recent one – scarcely a disaster, but it was not a life or death system.
We were not unique. Across the industry, people were conducting similar checks and corrections, with greater care and urgency wherever safety was involved. Incidents where routines would be disrupted or essential parts not serviced or replaced were caught before they could happen. The date changed without major incident and the workload returned to normal. We got accustomed to jokes about needing more stonemasons to cope with the year 1000 or reviving cryogenically preserved COBOL programmers in 9999. I still recoil from any year expressed with only two digits, but for many people there was nothing visible more serious than spelling ‘millennium’ without the double letters or celebrating at the start of the century’s last year rather than at its end. Now we have Dr Reid and his peers to thank that, saved from any ill consequences, we can pretend it was all a giant, humorous hoax.