For this platform I have used until now this simple scheme. During a initial phase, the global representation of each object is computed once for all and stored in compressed format (JPEG) on disc. When later the user selects a range of (ordered) objects for animation, these graphical representations are loaded in memory and an animation is started by displaying them one after the other. Limiting the size of the global representation to 100x100x256 allows for a scanning at the rate of 50 objects per second, giving a smooth animation. With this size also the overhead in terms of disc space is not so big (30000 objects occupy around 100 Mbytes of disc space).