[en] We point out an analogy between the Mattis ''spin-glass'' model and an electron model with random distribution of bounding and antibounding bounds. For a certain class of distributions, namely, those corresponding to zero gauge field, we use the Mattis method in order to perform a gauge transformation leading to the evaluation of all interesting characteristics of the system. The general result is a necessary condition for a system to be a random one is to contain frustrated plaquettes. Put in other words, real disorder is only gauge invariant disorder. (author)