The radiosity technique for the simulation of energy exchanges in synthetic models has been continuously improved in the past 15 years, to the point where impressively realistic images can be produced, and the technology is being offered in commercial products. Does this mean it can be used in practice? Many obstacles remain, obviously, for non-expert users who may want to produce meaningful simulations, and radiosity can hardly be described as a mature technology. We consider here the potential applications where a user will want to use radiosity: we identify the remaining technological barriers, and present some avenues being explored to lift them.