Paying a huge price during compression. I think the compression speed is yet to be vastly improved, for example there are no arch-specific implementations, like they did with decompression.