If you're implementing it for Computercraft anyway, there's no reason to stick to the standard. It's well known that bzip2 has a couple of extra steps which don

If you're implementing it for Computercraft anyway, there's no reason to stick to the standard. It's well known that bzip2 has a couple of extra steps which don't improve compression ratio at all. I suggest implementing Scott's Bijective Burrows-Wheeler variant on bits rather than bytes, and do bijective run-length encoding of the resulting string. It's not exactly on the "pareto frontier", but it's fun!