
Originally Posted by
ImaginaryHuman
Random data does have some redundancy to it. Byte and bit patterns do reoccur quite often, although in scattered positions. The problem is that typically those reoccurrence are spaced, on average, just slightly too far apart to be able to encode them in less bits. The trick then might be in simply finding an encoding method which is leaner and more efficient than the output from previous compressors. After all, if e.g. we ZIP a file and it gets to 90kb, and the output seems random, but then we LZMA a file it gets to 70kb, obviously both seem random but the output from lzma has identified an even more efficient way to encode the redundancy. And then there are other compressors which can also produce even greater compression. So although there may theoretically be a limit to such compression, it still stands that inefficiencies in a given compressor leave room for improvement. Data can look random and yet still exhibit some limited `potential` for further reduction, up to a point.