Are OpenGL calculations regarding dynamic lighting dependent on (or affected by) the camera? ie Does moving the camera require OpenGL to recalculate all the dynamic lighting equations?

And are there any solutions which fall between the computationally intensive per-frame lighting calculations versus the size-intensive per-polygon lightmap? I’m more limited on space/bandwidth than I am on computational power… I’m looking for as many values that I can precalculate as possible (ie polygon normals, vertex normals, etc, which I already do) to help the lighting system along without taking more than a few extra bytes (~50 bytes) per polygon.

Thanks!