INTRODUCTIONOn the macroscopical level, Einstein's general relativity has passed every test "with flying colors", see Will [67] for a recent review. However, Einstein's theory sofar has resisted any attempt of quantization, e.g. it is known to be perturbatively nonrenormalizable, partially due to its dimensional coupling constant 2 . String theory or brane scenarios with extra dimensions have been proposed as a rescue, some of which are implying, however, deviations of standard gravity in the sub-millimeter range. Recent torsion balance experiments [28] have probed the inverse square law and found no deviation even below the hypothetical dark energy (DE) scale of λ DE = (hc/ρ DE ) 1/4 ≃ 85 µm. Thus this 'window' of possibly new gravitational physics seems also to be closing.In 1974 Yang proposed an affine gauge theory gravity [70] which, due to its scale invariance, can be regarded as a rather promising fundamental theory of (quantum) gravity in the high-energy limit [25], without invoking extra dimensions or supersymmetry, cf. Ref. [30]. An additional duality constraint on the curvature could be extremely important for the path integral approach to quantum gravity. Then instanton type configurations [22] near the classical ones, i.e. Einstein spaces, are more probable then the 'spurious' Thomson spaces, as one would expect naively. For the modified duality with a breaking of scale invariance, the transition amplitude peaks at classical Einstein spaces only. Alternatively, in a four-dimensional Yang-Mills theory gauging the de Sitter group [37,54], scale invariance gets spontaneously broken by a pseudo-Goldstone type 'radius vector' [62], odd under CP, in order to recover the Hilbert-Einstein action plus the Euler term.From the work of Stelle [59] we know that the curvature squared gravity in Riemannian spacetime is perturbatively renormalizable but unfortunately plagued with physical ghost, i.e. negative residues in the graviton propagator [36]. This finding has diminished 1 E-mail: ekke@xanum.uam.mx 2 It has to be kept in mind that Newton's gravitational constant G is one of the less precise known constants of physics. In order to improve this situation, there are plans [1] to measure the gravitational attraction of two bodies in a spaceship (Project SEE), where the larger body will function as a shepherd for the movement of the test mass, similarly as in the rings of Saturn. Amongst others, an accuracy of 3.3 × 10 −7 in the determination of G should be feasible. the initial interest [19,20] in such models.Much more promising and elegant is to start from a purely topological classical action, proportional to the gravitational Pointrjagin (or Euler) invariant and then quantize this model by nilpotent Becchi-Rouet-Stora-Tyutin (BRST) transformations generated by s. Such a topological action is not only completely metric-free, but moreover conformally invariant [69,34] and can provide a consistent Topological Quantum Field Theory (TQFT) as shown by Witten [68,9]. Lateron, it was realized by Baulieu and Singer [6]...