float to bfloat .. disagreement meta-question

This may not be the right issue, but we may be faced in the
future with differences where groups want to go different ways
with the design or implementation.

How to resolve disagreements?

1. argue until reaching consensus (Common Lisp committee). If
no consensus?
2. break off in a fit of pique and form another language (Unix
standard behavior with respect to operating systems).
3. hide behind some switch (e.g.  
try_to_rationalize_float_before_conversion_to_bigfloat: true)
4. vote; majority rules.
5. appeal to authority.
6. standards "committee" authority.
7. Something else. (?)

I would think that (2) is bad; (3) is tiresome and could clog the 
documentation
and make interactions between different programs very touchy.  It is 
almost what
was done in the Macsyma design except when (5) could be made to work.

RJF


PS, with regard to decoding floats, I think that treatment of  reserved 
operands by integer-decode-float
is problematical.  Allegro CL which tends to be fairly standard's 
conscious, gives an error as shown below..

 (integer-decode-float (expt 2.0 1000))  ;; similarly for NaNs.
Error: Attempt to decode a reserved floating point value with exponent 
128 to an integer
  [condition type: simple-error]

I don't know what you'd like to do for bfloat(NaN) or bfloat(infinity)  :)