Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!lll-winken!uunet!abvax!iccgcc.decnet.ab.com!herrickd
From: herrickd@iccgcc.decnet.ab.com
Newsgroups: comp.arch
Subject: Re: IEEE arithmetic (Goldberg paper)
Message-ID: <1991Jun17.130828.4892@iccgcc.decnet.ab.com>
Date: 17 Jun 91 18:08:28 GMT
References: <1991Jun11.175639.22558@zoo.toronto.edu> <196@armltd.uucp>
Distribution: comp
Lines: 17

In article <196@armltd.uucp>, dseal (David Seal) writes:
> A fair point, and I agree it's open to interpretation. I would favour (*not*
> "insist on") the 'subset of the unextended real numbers' interpretation
> because (a) the unextended real numbers are a mathematical system which is
> *comparatively* easy to work with; 

One of the problems with this conception (and it dates back to Backus'
team) is that any system of floating point numbers (ignoring new fangled
ideas like nan and inf) is just a finite set of rationals.  The REALs
have odd features like
    1)  When you add two of them together, you get another of them
    2)  Addition is associative
while the floating point number systems don't have such properties
(because the underlying set is finite).

dan herrick
herrickd@iccgcc.decnet.ab.com