Why do you think so? The code example shows that you can do RLE (run length encoding) without noise / additional space. I'm pretty sure you can do zip as well. It would just be very hard to implement, but it wouldn't necessarily require that the output contains noise.