True. But there you can go into optimizations like parallelization of the compression and building an optimized dictionary upfront once to make the process faster.