> Why not just allow an implementation to "inline" substring?
> Instead of:
>         (string->number (XS>< "$12345.99" 1))
> the programmer would write:
>         (string->number (substring "$12345.99" 1))
> and the implementation can re-write this as:
>         (%%substring->number "$12345.99" 1)

I agree with Per.  It will be harder work for the compiler but the interface
is much cleaner.

Marc