Yes, you''re right, I have to use fragment programs or register combiners or something to achieve per-pixel lighting.
I just thought about writing my own lighting class with support for per-vertex and per-pixel lighting, visibility determination and depth-testing for lights and the possibility to add as many lights as needed and even if these would be more than 8.
I read some parts of Carmack''s .plan and he wrote that with the NV30, for example, there is most time only one rendering pass required in Doom III. He was talking about the vertex-programs he implemented and what extensions they use, so I don''t know if he meant there is only one pass for the vertex stuff required or one pass even for the per-pixel lighting he uses. So I thought ppl could give me their ideas, how they would implement lighting - that''s what I want