```On Sun, Jul 29, 2001 at 09:58:02PM +0900, Kero van Gelder wrote:
> 12345678901234567890 / 1.0  => 1.23456789e+19
>
> which actually looses accuracy, by automatically converting to float,
> instead of BigNum. This would be an argument to drop automatic
> conversion for floats. You could still convert automatically with
> Numerics and Rationals, since you loose nothing (and should expect a
> Rational anyway).

I like the idea of built-in rationals.

Since the difference between rationals and ints is much smaller than between
rations and floats, I believe fully automatic conversion between ints
and rationals makes more sense than between ints(rationals) and reals.

Rationals of course should automatically convert themselfes to Fixnums (or
Bignums) if possible.

e.g
1/2 = Rational(1,2)
Rational(1,2) * 2 = Fixnum(1)

For floats I am unsure what 1/2.0 should give. Since the /-operator is a
member function of Int in that case, one could argue that the result should
be 1 (or 0?). This would mean that "float on the left side gives float for
the result".

For division this _could_ be ok, but for multiplication its crazy.

1.0 * 12345678987654321 would give a float, thus losing accuracy, but
12345678987654321 * 1.0 would give an int.

Maybe there this be an extra set of operators for float?
kind of like: /., *.

greetings, Florian Pflug
```