Bifurcated structures are ubiquitous in nature, both in living and non-living systems. The modern approach to their physical modelling starts from the recognition that in plants and animals the substitution of (a portion of) a straight vessel length with a branched one serves a biological "goal" that requires some energy and material "expenditure". Such expenditure must be justified by a compensating gain for the resulting evolved structure. A great number of studies addressed this topic, mainly from a biomedical science perspective, but since all of them indicate a weak agreement with experimental data, it is useful to explore the matter in some more detail. The purpose of the vessels considered in this study is to "transport" material flows like sap, blood or air, and the striking geometrical similarity of the forked structures found in plants, circulatory systems, bronchial alveoli and river deltas suggests indeed the presence of a single underlying physical principle. Experimental evidence indicates that the topology of bifurcated blood, air and sap vessels (e.g., their "shape") is amazingly similar under quite different external constraints, and this might imply that the shape of a bifurcation is at least to some extent independent of the boundary conditions. Furthermore, it is unclear if and to what measure the functional advantage obtained by repeated bifurcations decreases with the number of splitting levels. This paper presents a critical review of the most popular physical model formulated for the description of bifurcated structures, the so-called Hess-Murray law ("H-M" in the following). It is first shown that, under a very restrictive set of assumptions, the H-M law can be obtained by the assumption of constant wall stress in the parent and daughter branches. Then both Hess' and Murray's original derivations are discussed from a physical point of view, and the extension of the rule to the case of non-symmetrical bi-and trifurcations is presented. It is then argued that in real branched networks the actual optimality criteria (i.e., in an evolutionary sense the "driving force") may be quite different from the assumptions posited in the H-M law literature, which explains the weak predictive value of the law. Some adjustments to the model that include a resource-based cost/benefit of the formation of a bifurcation are presented and discussed.