This stuff does put me to sleep, except on a marketing level; remember what
happened to Intel on the Pentium FP bug? "Oh, the typical user will never
run into this FP error - don't worry, be happy." One quarter later, they
were forced into a recall, and took a ~$500 million dollar charge against
earnings to cover it. Ouch.

We're not on the line for amounts in that order of magnitude, but we would
be remiss if we did not document the limits of our implementations
somewhere prominently (perhaps an updated math(3)? It's dated 5/6/91) so
that people who care can understand precisely what they're getting.

I just want the right answer from dc and bc when I do simple arithmetic.

	just a thought,

	Erik <fair@clock.org>