Le 23/04/2020 à 07:22, Michael Phillips a écrit : > The truly "fun" part about Y2K was that IBM solved the problem in the early > 60s with just 6 digits. CFO-64 was a life insurance application they wrote in > Autocoder that was my first encounter with what EDS-ers called "mo-year" > code. Dates were stored as a 4 digit number of months since some epoch > (sorry, I don't remember the actual epoch month/year...) and a 2 digit day of > the month. Some how or another EDS "acquired" the source and ported it to BAL > as LMS. IBM wasn't actually worried with Y2K then, they were just taking care > of policies for folks born before 1900! > > With 9,999 available months, that code was fine for well beyond Y2K, which by > the way, is actually still 28 years away... :)
That was supposed to fit in 3 bytes I suppose ? in which case, the even better solution was to count a number of days since some epoch. 100k days is almost 274 years if you go to binary numbers, 2^24 days, that's 45k years Raphaël ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
