We investigate the discrepancy principle for choosing smoothing parameters for kernel density estimation. The method is based on the distance between the empirical and estimated distribution functions. We prove some new positive and negative results on L_1-consistency of kernel estimators with bandwidths chosen using the discrepancy principle. Consistency crucially depends on a rather weak Hölder condition on the distribution function. We also unify and extend previous results on the behaviour of the chosen bandwidth under more strict smoothness assumptions. Furthermore, we compare the discrepancy principle to standard methods in a simulation study. Surprisingly, some of the proposals work reasonably well over a large set of different densities and sample sizes, and the performance of the methods at least up to n=2500 can be quite different from their asymptotic behavior.