RE: Representation of floating point values.

>Precisely.  The question is what to do about potential lexical
>representations whose exact value is too large.  float and double
>round almost all values quietly so as to get a value in the value
>space.  

Rounding is dealt with by the IEEE754 standard, and all applications
using it should know what to expect.  Similar issues with gradual
underflow.

I am asking not for an exception, but to follow an external standard
that Schema cites consistently - including support for traps and flags.

Daniel;