In porous media as elsewhere, models based on self-similarity of topology/geometry have received increasing attention in the last 20 years, especially in view of physical fracture treatments, which generate fractal media, and both experimentally determined systematic trends of density with volume, and established power laws for water retention, compatible with fractal geometry. Recent treatments of fairly large soil data sets by Hunt and Gee, 2002a, and of Filgueira et al., 2000, have shown that particle-size data can be used to predict the water retention characteristics of soils over a range of moisture contents which is typically about 1/6 of the porosity to near saturation, but doubt has been expressed regarding the usefulness/appropriateness of fractal representations on account of the deviations from prediction, especially at low moisture content. However, in Hunt and Gee, 2002a, it is shown that the moisture content at which this deviation occurs is typically the same moisture content at which solute diffusion vanishes (in a third soil data set) Moldrup et al., 2001, and is therefore best interpreted as the critical volume content for percolation of capillary flow. Using percolation theory to calculate the hydraulic conductivity of unsaturated flow, the fractal dimensionality of the pore space, the porosity, the saturated hydraulic conductivity, and the saturation (all experimentally determined) it is then possible to predict to a high degree of accuracy the unsaturated hydraulic conductivity, Hunt and Gee, 2002b, without unknown parameters. The knowledge that (probably) a large fraction of the world's soils are fractal, and that percolation theoretical treatments explain water retention, hydraulic conductivity, and diffusion relationships for these media, provides motivation for considering new theoretical treatments of the vadose zone in land-surface inputs to global circulation models. Although issues of scale, "geological" heterogeneity and coverage remain, it is suggested that a physically-based model with the most advanced treatment of pore-scale heterogeneity should be an optimal starting point.