Deprit's reduction of the nodes revisited