The notion of policy is widely used to enable a flexible control of many systems: access control, privacy, accountability, data base, service, contract , network configuration, and so on. One important feature is to be able to check these policies against contradictions before the enforcement step. This is the problem of the conflict detection which can be done at different steps and with different approaches. This paper presents a review of the principles for conflict detection in related security policy languages. The policy languages, the notions of conflict and the means to detect conflicts are various, hence it is difficult to compare the different principles. We propose an analysis and a comparison of the five static detection principles we found in reviewing more than forty papers of the literature. To make the comparison easier we develop a logical model with four syntactic types of systems covering most of the literature examples. We provide a semantic classification of the conflict notions and thus, we are able to relate the detection principles, the syntactic types and the semantic classification. Our comparison shows the exact link between logical consistency and the conflict notions, and that some detection principles are subject to weaknesses if not used with the right conditions.