I've had a few ideas kicking about and thought I'd throw them out there to see if anyone can smash them to bits with forseen problems :).
Could shadows be done using a render to texture effect where the camera used to capture the image to render is situated at the light source? I'm thinking the camera is pointed toward an object, it only sees that object and it renders it without light sources and with a black texture. It renders this to a texture applied to a quad with a percentage alpha blend (with white areas being transparant) which is moved to a position relative to the object casting the shadow depending on the position of the light source and is stretched to fit. Basically, each object would carry around a separate shadow quad for every light source in range. Each light source would have a camera which cycles through all objects in range which cast shadows and renders their shadow to the quad designated for that light source.
Or is the current way of doing shadows greatly more efficient than this? I guess you could pick the vertexes and draw a shape on the second opaque object in the picking order (the first being the one casting the shadow).
I'm an absolute newbie at all of this, just throwing a few ideas about.
What you're talking about is similar to Shadow Mapping. It's something we will add later most likely. Let me just say that there is no shadow technique that is right for every application. Some games/apps will benefit from using shadow volumes and some from shadow mapping. They both have pros and cons. We are still in our infancy stage for shadows though, so all the feedback we can get will help us move in the right direction.
Ah, righto. I have a tendancy to invent things that have already been invented ^^;.