Logged In: YES 
user_id=6380

Hm.

In 2.2a2, currently, pow(a, b, c) for ints a, b, c where b <
0 is defined as pow(float(a), float(b), float(c)), IOW
(1.0/(a**b))%c. This doesn't make a lot of sense for the
three-argument version though, because the result tends to
be between 0.0 and 1.0...  But it is consistent with the
(future) rule that operations on integers and floats should
give results with the same value only a different type.

Assigning to Tim, whose mathematical sensibilities are more
refined than mine...