Bulk glass formation occurs over a very small part of phase space, and 'good' glasses (which form even at low quench rates similar to 10 K/sec) select an even smaller part of that accessible phase space. An axiomatic theory provides the physical basis of glass formation, and identifies these sweet spots of glass formation with existence of rigid but stress-free networks for which experimental evidence is rapidly emerging. Recently, theory and experiment have come together to show that these sweet spots of glass formation occur over a range of chemical compositions identified as 'intermediate phases' (IPs). These ranges appear to be controlled by elements of local and medium-range molecular structures that form isostatically rigid networks. IP glasses possess nonhysteretic glass transitions (T-g's) that do not age much. Raman scattering has played a pivotal role in elucidating the molecular structure of glasses in general, and in identifying domains of IPs. Experiments reveal that these phases possess sharp phase boundaries and are characterized by an optical elasticity that varies with network mean coordination number, r, as power law. In this review, we provide examples in chalcogenide and oxide glass systems in which these phases along with optical elasticity power laws have been established. IP glasses represent self-organized nanostructured functional materials optimized by nature. Copyright (c) 2007 John Wiley & Sons, Ltd.