I believe the below is worth reposting in its entirety...
From...
http://mathforum.org/library/drmath/view/55021.htmlDate: 06/26/98 at 16:25:15
From: Doctor Barrus
Subject: Re: Is pi a constant in non-Euclidean geometry?
Hi, Sean!
Well, you're both right in different cases. Whenever you're talking
about perfect circles on flat (Euclidean) surfaces, then pi, which
represents the ratio of the circumference of the circle to its
diameter, is a constant. (By the way, this number is irrational but it
IS a constant, just as the square root of 2 is a single irrational
constant whose decimal expansion goes on forever. Take 1, 1.4, 1.41,
1.414, or any other approximation of the square root of 2 and square
it, and you won't get 2. Only the irrational square root of 2 squares
to 2. Similarly, there is only one number, pi, in all its irrational
glory, that represents the circle's ratio.)
However, if you go into NON-Euclidean geometry, where you deal with
curved surfaces, then the ratio of a circle's circumference to its
diameter does not remain constant. For example, say you stretched a
piece of rubber over a circular hoop. When the rubber lies flat,
you've got a flat (Euclidean) circle, and the ratio of the
circumference to the diameter is pi = 3.14159265358... But say you
poked your finger through the center of the circle and stretched the
rubber a bit. Then the diameter of the circle would grow, but the
circumference would be the same. The ratio would change - it wouldn't
be constant. Consequently, if you called this ratio pi, then pi
wouldn't be constant.
It gets a whole lot more complicated in non-Euclidan geometry than
this. For example, suppose you traced a circle (and you have to be
careful of your definition of circles in non-Euclidean geometry) on a
horse's saddle. This constitutes a surface with negative curvature.
The ratio of circumference to diameter is not easy to compute exactly
without advanced mathematics, let alone to describe!
So you see, you're right that pi is a constant when you're dealing
with Euclidean, or flat surfaces. However, your friends are right in
that the ratio of the circumference of a circle to its diameter
changes and is not constant in non-Euclidean spaces.
Whew! Long answer, but I hope it's helped.

- Doctor Barrus, The Math Forum