Friends, some days I simply don't know what we're doing. Open your nearest JavaScript console (or whatever) and observe

350680 / 8813712 * 2203428 == 2203428 * 350680 / 8813712

. I'm sure you all know why this is false. But what can we do in the face of such sandy foundations?

Better yet: 0.1 + 0.2 + 0.3 - 0.3 - 0.2 != 0.1

Do like Woz and encode all numbers as fixed precision.

IEEE floating point leaves a lot to be desired. We are planning to support Crockford's proposed DEC64 arithmetic, which solves the 0.1 + 0.2 = 0.3 problem. Obviously has huge implications, as the underlying OS uses IEEE floating point.

JS is definitely the sandiest of foundations.. there are better ones. But yes, no escaping IEEE when it is what the hardware does

Having turned it over a bit more, one of my frustrations is that Aspect Oriented Programming didn't become a standard thing because I would like to be able to tell the compiler, "Okay guys, within the important financial calculation package, let's not use doubles, kthx."