RE: Strange result with exponents...

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: Simon Coulter
> 
> The ** operator has used floating-point since its inception.

That's been made clear now, which makes it even stranger that the
results have changed from one release to another.

 
> The only 'error' was choosing to use floating-point when both operands
> were decimal.

I actually never used the word "error".  I simply said that if things
work differently between releases, this needs to be pointed out.  I
don't care if it's because something got "fixed"; if behavior of a given
opcode changes between releases, there should be some notification.

This particular situation, however, may be potentially less of an issue,
since as you say, "anyone who uses floating point should know that it is
an approximation and is only accurate to nearly 16 digits."  At the same
time, should everyone who uses exponentiation know floating point?  I
don't think that's reasonable.  As you say, it would probably have been
a better solution to use integer math when both operands are integers.

Anyway, my point is still this: backwards compatibility has been until
recently a mantra at IBM, and I hope it continues to be one.  Because of
the nature of this specific issue, however, it is not a good measurement
of whether or not it's the case.  Someone could reasonably argue that
what happens after the 16th digit need not pass the compatibility test.

Joe