An example is given for which the diagnosing algorithm for networks, proposed by T.Kameda, S.Toida and F.J.Allan, fails to compute the correct system diagnosis. A correction of the algorithm is described.