This is apparent. xz's own game is not "a specialized compression pre-processor for x86_64 ELF binaries.". xz's own game is a general-purpose compression utility suited for a range of tasks, not optimized for one ridiculously specific domain. Also, any compression benchmark really ought to include speed of de/compression, not only compression ratio, as compression algorithms occupy along a scale trying to maximize one trade-off or another.
I never claimed to beat xz as a general-purpose compressor. .tar.xz is the dominant format for Linux source tarballs and distro packages. So optimizing for ELF + x86_64 is optimizing for a very real and common case, not some toy benchmark.
btw goal of the project was not building a production ready solution. It was curious case of black box software development. Compression is great because input and output are precise bits. As for speed, I think it's comparable since it's using most of XZ infra anyways.
This is apparent. xz's own game is not "a specialized compression pre-processor for x86_64 ELF binaries.". xz's own game is a general-purpose compression utility suited for a range of tasks, not optimized for one ridiculously specific domain. Also, any compression benchmark really ought to include speed of de/compression, not only compression ratio, as compression algorithms occupy along a scale trying to maximize one trade-off or another.