See the file http://www.cs.berkeley.edu/~fateman/temp/decode-floats.lisp
If this works (compiled) in other lisps, it might solve the problem of
decoding nans and infs in a uniform fashion. I sent Robert a version of
this file but it seems that it was corrupted; the one posted should be
better.
On the to-do list: run it on GCL, SBCL, CLISP etc.
It runs on Allegro.
RJF