The industrial importance of adhesives is constantly increasing. Yet it is difficult to systematize the vast amount of practical knowledge, which has accumulated, covering chemistry, interfacial physics, and mechanics. This review describes an attempt to bridge the gap between polymer science and fracture mechanics. It focuses on weak mechanical junctions. Examples can be found at glass–rubber interfaces or at glass–plastic interfaces, where the glass has been grafted with polymer chains that promote adhesion. When a fracture propagates along such a junction, the dissipation tends to be localized in the junction region. We present a phenomenological description of this process in terms of two ingredients: (i) a threshold stress σc associated with chemical scission or with plastic flow; (ii) a "suction" process with a suction velocity proportional to the local stress σ, which ends when the volume transfer (per unit area) has reached a certain limit hf. Assuming no cavitation (no crazes), we are led to expect two fracture regimes: (a) at low-fracture velocities V, the process is quasi-static and the fracture energy G scales like σchf and (b) beyond a velocity V*, the width of the suction region is very much spread out, and G increases linearly with V. On the whole, these ideas can put into perspective a number of existing data, for instance, we may understand why adhesive elastomers become poorer when their level of cross linking is increased.