predictable behavior and optimization setting, was: +-Inf and NaN

[Robert Dodier]

Camm, thanks for taking the time to write about this stuff. I'm more
broadly interested in all of it, but in the interest of brevity I'll
just make a quick note.

It sounds like part of the proposal is to let program behavior change
with the compiler optimization setting. I'm pretty uncomfortable with
that -- it would make it pretty hard to predict what some code is
going to do. There's no guarantee that the optimization flag is the
same one time as another, so one can't know whether the program will
behave the same.

I'll let that be enough for now and try to address some of the other
stuff later.

best

Robert