[en] From both the theoretical and the practical points of view, the problem of constitutive laws is a part and parcel of the modeling problem. In particular, the necessity to restore in the model, through topological laws, some of the information lost during the usual averaging process is emphasized. It is shown that the customary 'void fraction' topological law Psub(V)=Psub(L) should be proscribed whenever propagation phenomena are involved. A new void fraction topological law is proposed. The limitations of the current assumption of constant pressure within any phase in any cross section are also illustrated. The importance of proximity effects (neighborhood and history effects, related to characteristic lengths and times) is brought out. It results in the importance of the mathematical form of the constitutive laws. Various approaches to the constitutive law problem and possible mathematical forms for the transfer laws are reviewed. The simplest form (transfert terms as functions of the dependent variables only) may have some usefulness if interpretation of the results in terms of propagation phenomena is banned. A good compromise between the necessity to take proximity effects into account and to obtain a tractable set of equations is carried out when so called 'differential terms' are introduced in the transfer laws. The last part of the paper is devoted to some restrictions, which are imposed to the transfer terms because of some basic principles: indifference to Galilean changes of frame and to some changes of origins, second law of thermodynamics and assumption of local thermodynamic equilibrium, closure constraints. Practical recommendations are formulated