Results on hard diffraction from CDF are reviewed and compared with predictions based on the diffractive structure function of the proton measured in deep inelastic scattering at HERA. The predictions are generally larger than the measured rates by a factor of {approximately} 6, suggesting a breakdown of conventional factorization. Correct predictions are obtained by scaling the rapidity gap probability distribution of the diffractive structure function to the total integrated gap probability. The scaling of the gap probability is traced back to the pomeron flux renormalization hypothesis, which was introduced to unitarize the soft diffraction amplitude.