Our understanding of dry-snow slab avalanche release improved over the last decade - not the least by consistently following a fracture mechanical approach. Whether we consider artificial triggering or natural release, slab avalanches result from a sequence of fracture processes including (i) failure initiation in a weak layer underlying a cohesive snow slab, (ii) the onset of crack propagation, (iii) dynamic crack propagation through the weak layer across the slope, and (iv) tensile failure – equivalent to crack arrest, followed by sliding of the slab. While failure initiation is best understood in terms of applied stress and strength, crack propagation is best understood in terms of stress intensity and fracture toughness. The typical anisotropic microstructure of persistent weak layers favors failure in shear rather than compression under mixed-mode loading - though the failure type at the micro-scale is largely unknown due the complex stress state in the ice matrix. The fracture mechanical approach has also changed our view on the spatially variable nature of the snow cover. Spatial variations of weak layer as well as slab properties may control avalanche formation, since disorder is fundamental for the fracture process. For example, failures will initiate from locally weaker spots, and fractures may arrest due to locally stronger areas. Whereas we still lack a comprehensive model linking damage at the micro-scale to avalanche size, recent modelling approaches have demonstrated - by assuming realistic failure behavior of the weak layer including its collapse and resulting mixed stress states in the slab layers - that not only failure initiation, but also crack propagation depends on slope angle. We present a modern synthesis of avalanche release.