on a typical record length of 16 byte:

bin size = hex size / 2.8


I use float on many applications on the AT89C2051 without any problems.
E.g.:
- read something in from a sensor
- do some corrections x = A * x + B or so
- convert to digits and display it on LED / LCD

If speed was not an issue and applications are small and writing the code was many times easier, why not.


Peter