That same year, 1982, I was a 2nd year university student. At the end of the year I got a summer job in my small home town (~40k) as the sole programmer doing a project to computerise the city's 30 cm thick binder of loans various individuals and institutions had made to them to build or upgrade the town water supply or library or whatever. Each loan had an interest payment due quarterly, half-yearly or yearly and every quarter the finance manager would leaf through the binder noting the payments due that quarter. Some terminals were available connected to a PR1ME minicomputer at a nearby bureau (which had essentially two customers: the city council, and a dairy factory). COBOL and FORTRAN were available, and an ISAM database.
Even in 1982, some of the loans matured a little into the 21st century, while others had issue dates as far back as the early 1950s. I suggested that we should store years with 4 digits but it was impressed on me that this would not be acceptable due not only to the wasted disk space (ha!), but also would not be compatible with standard data types and ad-hoc query software for the database.
I came up with a scheme where a single global parameter was stored, a two digit year, such that years in the database greater than or equal to the parameter were considered to be in the current century, while years less than it were considered to be in the next century. I initially set the parameter to "50", meaning that dates from 1950 to 2049 could be represented. Once loans from the 1950s had all matured they could be aged out of the system (there was a function for this), and the parameter increased to "60" etc. This was also done automatically by the ageing function.
And so in 1982, at 19 years old, working alone, I avoided the Y2K problem! I'm a little bit chuffed about that :-)
[There was another rather interesting problem in this system ... some of the debenture certificates listed explicitly the date and the amount of every payment. I of course didn't want to store every payment, but calculate it from the principal and interest rate. However some of the printed numbers (in particular from the AMP insurance company) were one cent different from what calculation would suggest. It was also impressed on me that if an incorrect payment was made the AMP was actually stupid enough to send an invoice or refund for 1c and we didn't want that to happen. I eventually solved the problem, after noticing that the problem debentures had all been issued before 10 July 1967. A chocolate fish to anyone who can tell me why, and how to fix the calculation :-) :-) ]
Home
Help
Search
Login
Register
