Changing the Division Operator -- PEP 238, rev 1.12

[Arthur Siegel]

Mark writes -

>Somewhere in one of the posts on the many PEP 238 threads was the
>following link to a discussion in a google group:
><http://groups.google.com/groups?hl=en&safe=off&th=1d9d3c24ba2f933d>

Thanks for that. I found that thread actually very interesting, and am sorry
I missed the cite the first time it came around in the PEP238 thread.

But what I found interesting, as a novice, was discussion and advise
from some of the Big Guys about doing mission (life) critical work
within the limitations of *any* implementation of floating
point numerics.

The "/" issue was a digression from the meat of the conversation
 - which was really about testing techniques, and in general
belts and suspenders when traversing the floating point minefield.

Tim writes -

>If you have a programmer who, e.g., assumes that x+y-x equals y, or thinks
>(-b + sqrt(b**2 - 4*a*c))/(2*a)
>is a good way to find a root of a quadratic equation, they'll eventually
>kill someone regardless of programming language.

And regardless of division sematics.

No?

That is what is trying to be tackled with Unified Numeric model.

Correct ?

A Big Thing.

Will be interesting to watch develop.

ART