After writing this paper I found that the result had been proved elsewhere. The paper might still be of some interest serving as an introduction to the classifaction of the complexity of constraint languages