Floats aren't the problem
Why are floats getting a bad rap? Every data type has its limits, and requires that you know how to use it properly to avoid errors. Other data types are better at certain types of calculations, but only if the programmer knows when to use them, and what the differences are between them.
Integers are great for counting things, but don't kid yourself that they solve your rounding problem in any way. They only make it much worse and totally out of your control. And if you move the decimal over enough that the always-truncate rounding method doesn't seem like such a problem, you had better watch your upper bound if you want to do any real math (like multiplying two values together).
BCD can help avoid rounding when it seems unnecessary in decimal, but even ignoring the computational overhead, it is easy to contrive situations where BCD rounding is signifigantly worse than float.
Data types are just tools. Only a very poor craftsman would use a screwdriver to drive a nail, or blames his hammer for not being good at cutting boards. Or rounds a float and crashes his satellite.
Biting the hand that feeds IT
