Because the cpio utility has awful usability

(it might capture more info than tar maybe? but it's not worth the effort)

Second that. I had the time when I was working with initrd images, so I had to use cpio a lot. I could never get used to its syntax.

And the fact that it’s manpage was missing, only saying “use info instead” did not help its usability at all.

That was a GNU documentation problem, never applicable to non-GNU versions of cpio, and not even applicable to the current GNU manual page, which has no mention of info and definitely exists. It also post-dates by years the period where tar and cpio were in competition, which was (roughly) the early 1980s. (The POSIX standard pax aimed at resolving the conflict was in the 1003.1b draft in 1991.) It thus is an entirely fallacious explanation for what happened in the "archiver wars".

* https://news.ycombinator.com/item?id=18978064

* https://netbsd.gw.com/cgi-bin/man-cgi?cpio

* http://osr507doc.sco.com/man/html.C/cpio.C.html

* https://manpages.debian.org/sid/cpio/cpio.1.en.html

Cpio did have some advantages. It was more space efficient, and would read past errors. See https://www.gnu.org/software/tar/manual/html_section/tar_71....

I think it was because of tape-drive manufacturers. Tar was often used to move big data off the platters and onto magnetic tape - cpio didn't have that killer application. In the bad old days of metal, tape was how you saved things, while the disks were for processing ..

I guess that once ftp sites and Usenet binary groups had de facto standardized on tar files (first compressed, later gzipped; uuencoded for Usenet), it was a foregone conclusion. Why they did, I don't know; it may have been random. Originally, lots of source was distributed as a shar file (also possibly compressed and uuencoded); maybe the name similarity with tar helped?

Maybe skip cpio and go directly to dar?

https://en.wikipedia.org/wiki/Dar_(disk_archiver)

I started using Linux in 1993 and cpio was barely used then.

maybe it was preference at first, some people finding it annoying to deal with cpio's file list, and it built up to the modern stuff like builtin integration with compressors (probably coinciding with tar.gz becoming the de facto *nix archive)? I personally find xz and cpio both cool, despite both being barely used ;)

XZ used to be the compression format for many Linux packages in many distros (at least Arch based) and has only lately been replaced by ZStd. How do you come to the conclusion that xz is barely used?

I definitely would not say zstd has replaced xz, but it is on its way too. I'm not sure it is even the best replacement because the needs of packages differ from the strengths of zstd. Zstd is designed to optimize for both compression and decompression (i.e. It does neither spectacularly but has very good/pareto trade offs) while packages are very much written once and read many times. If xz weren't so memory intensive to decompress it would be a good choice (ie it's not a good choice), but I feel like there is still a better option out there that approaches the problem from a "optimize for size at the expense of heavy compression costs" and zstd is not it

tar is far easier to use, I think.