While the LoS aspect is potentially desirable in specific applications, I agree that initially it can be perceived as a limiting factor. Researchers at Boston University,Rensselaer Polytechnic Institute, and Nakagawa Lab in Japan are working with software tools to characterize non LoS applications using reflective and refractive surfaces found indoors. A powerful system can be envisioned if someone managed to utilize these principles with a device that contains a receiver with a wide view.

Or you have a "networked" light socket in each room and a discovery protocol to figure out what room to route the data to. More work on the initial infrastructure side, but then you don't have to worry about mesh- your room is covered.