I realize this may have been mentioned, and apologize in advance.
Old news.  Anyway so ruby floats are *very* subject to rounding errors which bite programmers often.  
I move for the adoption of 'BigDecimal' (or 'Longdecimal' [from rubyforge]) as the default type for precision, and then
let programmers choose to use floats later if they want to optimize for speed.  This would save much headache to programmers,
and allow them to worry about speed optimization later.  Thank you!
-Roger