David Hall is a y2k programmer. He warns that testing before 2000 will not prove what programmers hope, namely, that their repairs will work in 2000.
This appeared on Peter de Jager's discussion forum.
* * * * * * * *
This is true for all systems, be they mainframe, midrange, PC, or embedded. I don't care how much you test, only the actual event (in this case 01/01/00) will show you how your systems will operate and if they will fail. I'm sure there will be numerous people holding their collective breaths throughout 1999 and 2000. Regardless of test results.
How many times have we built and tested systems, only to have them screw up during actual use? Or after months of use when one special logic path was finally implemented? Does anyone believe that Year 2000 development will be any different? Tests CANNOT substitute for real operation under real conditions with real events. They can at best enable you to eliminate obvious errors and problems. The rest is based on hard work and luck.
This makes contingency planning and continuous management contact doubly important as an ongoing part of any Year 2000 program.