Re: Issue FIXNUM-NON-PORTABLE (Version 3)

[Dan L. Pierson <pierson%mist@multimax.ARPA>]

    I'm willing to split the issue into two parts, one of which says: constrain
    FIXNUM and remove BIGNUM, and the other is to constrain FIXNUM and define
    BIGNUM to be exactly (AND INTEGER (NOT FIXNUM)).
    In the SUBTYPEP-TOO-VAGUE issue, we should probably make sure the
    requirements for FIXNUM and BIGNUM are there.
    
Here's why I don't like the second proposal:

Consider an implmentation with three numeric representations:

Fast                (INTEGER -1024 1023)
Immediate           29 bits
Extended            Multi-precision

(I understand from the discussion that similar implementations exist.)

Then, with the second proposal:

FIXNUM is Immediate
BIGNUM is (OR Fast Extended) because Fast can't be a FIXNUM.