I read Real-Time Rendering and the author explains that scene graphs can be used to propagate "materials" and light sources down the hierarchy, that's cool but it really ends right there. There's no explanation on methods and techniques of doing these stuff.
Right now I'm using scene nodes as containers for geometry objects and the farthest I've gone is doing skeletal animation with them.
My question is how such a system of using scene nodes as cameras, lights, shadows etc. can be be implemented. (A reading source will be gratefully accepted as well)