On Feb 26, 2004, at 9:20 PM, Amir 'CG' Caspi went on and on about: > > True, but they ARE accurate when you add base-2 numbers, which are > perfectly representable. 2+2=4 should work accurately on all binary > systems, even in floating-point. Should being the key word ;) An integer operation still would have done the job here. Whoever wrote the calculator program was lazy, and decided to float all the numbers. This kind of problem could have been avoided if they determined the difference between an integer and another number, and used the correct operation where needed. You make a good argument, nicely done. CJ