It looks like he wants written data to stay on disk until it is read out, which sockets wouldn't do. I don't feel comfortable with his approach though. Maybe I just don't "get" his problem.

edit: I was also thinking of mkfifo -- where does mksock exist? Never seen that command.

Gah, clearly I fail at life today! Yes, I meant mkfifo. Kind of pathetic of me... posting that from my Fedora 22 laptop not even <TAB> completing the name. It is an oldie, but a goodie.

Contents in a named pipe do not persist across OS restarts. Also a named pipe has limits about how much data you can put in it, because it's in-memory (http://man7.org/linux/man-pages/man7/fifo.7.html, http://man7.org/linux/man-pages/man7/pipe.7.html) The persistent "pipes"mentioned in the post have none of these limitations (data persists as if it was in a regular file and the only limitations on size are the one imposed by the filesystem)

SEJeff has nailed it [as he frequently seems to]!

It's not immediately clear to me why the author is referencing pipes at all rather than just file buffers or some other type of buffer.

No he didn't, see my answer to his post :)

Feel free to suggest a better name though...