Beside the (floating) point

Nice article that puts into "laymen"-esque terms one of the computational idiosyncrancies of binary arithmetic.

I've written about some of the fun you can have exploiting floating point formats in the past, but didn't delve as deeply (or clearly) into the precision issues as I might have liked - brevity is NOT something I'm known for :)

I'd suggest that numbers are like strings: given fixed precision, you should deal with them in their "native" base/format.