If you already have a parabolic mirror, you could use autocollimation. Set up the test as if you were testing the mirror, but use a known good mirror instead.
This will be more sensitive to local defects (rings, holes, turned edge) than to an overall spherical shape, as having a slight, even curvature would, I think, just move the focal point a little bit. Would it work, do you suppose, to detect this by holding the parabola and knife edge in place, and then change the position of the flat by a large amount? It seems to me that a shift in focal point would take place.
Mike_Crawford@QuickMail.Apple.Com