Beebe's comments on integral.txt
It is correct that the integral.txt proposal is consistent with
widespread language implementations and applications usage but not with
754 and C99's dynamic-rounding-modes and inexact-flag-laden integral
functions. That is because these were in my view useless and mistaken,
defined from a 1970's software implementation mentality, of which I was
a part, that held flag tests and
mode changes to be cheap compared to arithmetic operations.
I've expanded the rationale in integral.txt accordingly.