I'm not sure why raw mode would be particularly loading on a system unless it is
receiving a lot of data. In the situation described, it's doing nothing other than
waiting on the user. The load should be the same as if it is waiting in cooked mode.
Ah, those old HP terminals. I had one that had many pages of memory and would even
spool off the older stuff on to one of its two data tape drives.
The other ones we had too little memory. They couldn't even buffer 24 x 80
characters, so if you had long lines, the thing would start scrolling even before it hit
the bottom of the screen.
We had some terminals (I think VT100's had this option) that would automatically send
a ^S to the host when the screen filled and then you could release it.
I still remember the day I was working at RU. One of the young female operators says
"hey, someone is catting /usr/dict/words to my terminal." I told her to hit ^S
immediately to freeze the sending process and then ran a PS to find the culprit.
This reminds me of the dichotomy of "write" versions. The early ones did just
rely on being able to open the remote user's terminal. We found this rather
antisocial, so at JHU we always left the terminals protected against world write. We
used the x bit as the "want to receive writes" flag and made write setuid (this
also allowed us to avoid fun and games like people redirecting the input to things that
weren't typewriters).