$epsilon/epsilon$: three years later

Three years after the completion of the next-to-leading order calculation, the status of the theoretical estimates of $epsilon/epsilon$ is reviewed. In spite of the theoretical progress, the prediction of $epsilon/epsilon$ is still affected by a 100% theoretical error. In this paper the different sources of uncertainty are critically analysed and an updated estimate of $epsilon/epsilon$ is presented. Some theoretical implications of a value of $epsilon/epsilon$ definitely larger than $10^{-3}$ are also discussed.