Wikipedia says it's part of the wishlist Wordlist or Words package
On Debian and Ubuntu, the words file is provided by the wordlist package, or its provider packages wbritish, wamerican, etc. On Fedora and Arch, the words file is provided by the words package.
But from where is this compiled? I've been using it to 'help' solve crosswords at work, but there's a lot of words it is missing, and a lot that I don't recognise.
What are the origins of this file?
Over in BSDlandia you'll have to dig back through the 386bsd and possibly older sources:
http://cvsweb.netbsd.org/bsdweb.cgi/src/share/dict/?only_with_tag=MAIN
Looks like someone typed up words from "Webster's Second International" on account of the 1934 copyright elapsing according to this file:
If you follow the package descriptions to their source, that gives the information (many ultimate sources). For wordlist, that goes to SCOWL (And Friends), whose page lists other sources which are compiled into this collection. Tracing the contributor for a given word is likely to be difficult though.
Further reading:
I can't find a reference right now but I believe a lot of it comes from the Moby Project by Grady Ward
