http://www2.ati.com/developer/gdc/D3DTu ... hading.pdf
This article describes a technique that basically generates lightmaps of whose texels do not store the whole (cosine weighted) integral of incoming light, but three separate values, or the incoming light projected onto x/y/z. This must be done in tangent space.
The three lightmaps can be stored in separate textures or in separate channels of a single lightmap (without colored lightmaps).
It sounds a bit more confusing than it is, but the results are pretty impressive, and if L3DT goes on to support MegaNormalMaps (take that Carmack, I just coined a term!), for our textures, this would be a way to integrate that detail to our high resolution lightmaps as well.
Even the leading lightmapper, Giles does not support this functionality, so I would love to see L3DT one-up them, as well as use this feature in our own projects.
Even in it's suggested form, I wouldn't expect to come about in any near patch necessarily, but it's good to have solid, marketable, usable, and professional features (as L3DT has a great habit of doing) on that TODO list.
