Why would convexity matter when considering coverage?
I don't see how the interpenetration case is possible. By the time we reach this point in the rasterization pipeline, shouldn't all the triangles be 2d? So they shouldn't be able to intersect like that.
Would it ever make sense to have the pixel alternate colors for two/three triangles instead of computing an average of the colors?
What type of graphic would cause interpenetration of triangles to occur?
When would this interpenetration of triangles case occur and would doing the average of both or just having one of the two colors be the best solution to this?
Why would we have inpenetration if the triangles are 2D?
Could we just split the interpenetrated triangles into smaller triangle pieces that do not interpenetrate?
Could you give some specific examples on when will these cases happen?
How does lighting change the giving effects of interpenetration when we consider coverage?