That's only half the story though. Serial ports in UNIX aren't just telecommunications, they're session control, job control, raw interaction.

I believe this all stems from the way terminals used to be used... You would login through a terminal, and that terminal would belong to you. Any processes you created would know what your terminal was(it would be your "controlling termina"). The kernel would know which processes were allowed to control it(foreground processes) or not allowed (background processes). Hitting ctrl-C would send SIGINT to processes belonging to your terminal. When you logged out, the group of processes belonging to that terminal would be killed.

Logging in through a remote teletype was almost the same, since it was an extension of the serial port. Modems had a few more signals, UNIX would know when a modem hung up for instance.

Having a "controlling terminal" does several things. For one thing, it means that all programs you run will be able to find out what your terminal is(they inherit their controlling terminal from you), and get direct access to it if they want it (by opening /dev/tty). This means that, even when you bury ssh in a 9-deep pipe chain, it can still go directly to the source and ask for your password. Or if /dev/tty can't be opened, they know there's no human there to ask for a password and just give up. If you ever see an error like "No tty present and no askpass program specified", that's what it means.