Problems and solutions in the demo implementation

Screenshot 1 Uniform shadow mapping
Screenshot 2 LispSM shadow mapping in the same view as above produces bad results. the view vector is not the best choice for the up vector in this case, because most of the view frustum body is outside the intersection body B and therefor the view vector points in an suboptimal direction for our purpose of sampling the intersection body B from the light view.
Screenshot 3 LispSM with a better choosen up vector for the calculation of the shadow matrix gives good results again, because here we use the body vector of the intersection body B for the up vector, which correctly considers the greatly reduced view frustum.