@amerika @taylan @Suiseiseki @romin Yeah, xz underperforms in terms of output size: depending on what you're compressing, bz2 is smaller sometimes. xz is faster in the default configuration because it's parallelized, but the "p" in "pbzip2" is "parallel".

The relevant thing for me was that if I want the data to stay accessible, I want a compression algorithm that is old enough to have more than one implementation. The xz vulnerability demonstrated that the risk was real.
xz_meme_formats.png