By now you’ve probably seen this comic:

In 2001 the IEEE, promulgators of the POSIX standard, set out to fix this problem,¹ and came up with a solution: pax.

pax, like tar, is both the name of a file archive format and of a program used to create and unpack them; but the pax program, by default, produces and unpacks tar files, with a much simpler interface than tar: it’s quite easy to commit the basics of it to memory.

pax was standardized in the POSIX standard of 2001, and is included with most Unix variants; if it doesn’t come with yours, it’s very probably available in the package manager under the name pax.

Five-minute introduction to pax

tar talks about ‘creating’ and ‘extracting’ archives; pax talks about ‘writing’ and ‘reading’ them.² pax does not do file I/O to read or write an archive file itself: you always use it in a pipeline. With this in mind, let’s begin with an example:

pax -r < archive.tar

This invokes pax to read (-r) the file archive.tar and extract it to the current directory. The < is a Unix shell idiom for creating a pipe from a file to a program; it’s functionally equivalent to cat archive.tar | pax -r, which, apart from being longer to type, is also a classic Useless Use of Cat.

If a file is compressed with gzip, bzip2, xz, etc., you just add the decompression program to the pipeline — or use the (non-standard, but supported by most implementations) -z and -j options for gzip and bzip2, respectively:

gzip -d | pax -r < archive.tar.gz
pax -rz < archive.tar.gz

Here’s an example that shows how to create an archive:³

pax -w <<< directory > archive.tar

This tells pax to write (-w) an archive containing directory to the file archive.tar. Again, you can add compression programs to the chain, or use the -z and -j options:

pax <<< directory | gzip > archive.tar.gz
pax -wz <<< directory > archive.tar.gz

That’s about all you need to know in terms of making and unpacking archives with pax. It has a few more tricks up its sleeve, though.

You can list the names of the files in an archive by piping an archive to it without passing -r:

pax < archive.tar
gzip -d < archive.tar.gz | pax
pax -z < archive.tar.gz

You can also use pax to copy directory hierarchies by using -r and -w together. You could use cp -R for this, but there are some pitfalls to that approach. (My personal preference is to run a local-to-local rsync job, but pax is there, if you want it.)

pax -rw dest_dir <<< source_dir

You now know all you need to start using pax. The mental burden of all those fiddly tar options is now over for you. Go forth and archive!