This technique has become more of a standard post-production technique, as many second and third generation GIF editing packages have incorporated it as an auotmatic feature.
Planning ahead is not an integral part of this method, but I will include it here just to keep the concepts together.
In the "Flipbook" model imagine frames 1 and then 2 (examples A & B). The difference between frame 1 (A) and frame 2 (B) is that shown in C. It contains the new image information and position of the disc, and enough of the background material (to the left) to cover the previous image of the disc in frame 1 (A).
An exact difference would be a more complex shape that would exactly obliterate (mask) the disk image in frame 1 (A), and not bother with any of the background pixels common to both frames 1 & 2.
Many of the programs which do this differentiation don't bother with that precise of a difference, and simply slice the image vertically , producing a mask to cover the information that came before, plus the information that is new.