Teletype I-O interface

The multiplexor interface was designed to present each TTY as a separate device. The TTY was run in full duplex mode, with an echo for each typed character. This permitted a visual check that the character had actually been received by the computer. Wherever possible the echo was done by the PPU itself. Logically, this echo should be provided by the receiving process when it receives the character. The process can then do special purpose functions, such as echoing a different character than it received, or not echoing at all. These functions are useful when interpreting non printing control characters as special signals, and when receiving passwords. Also, the echoes can appear at appropriate places in the output of the process. It is desirable, if possible, to echo a character immediately after it is typed, since unexpected delays in the echoes are unnerving to a user at a teletype. Unfortunately, it is very expensive to permit a user process to echo each character as it is typed. The PPU had tables of 'break' characters, one table associated with each TTY. If a character arrived from the TTY that was not a break character it would be automatically echoed. If it was a break character, it would not be echoed, subsequent characters also would not, and a signal would be sent to the process involved. Furthermore, if a character arrived from a TTY while ordinary output was in progress on that TTY, the echo would be prevented and a signal for the first such character sent to the user process. Thus characters typed during output could be echoed by the receiving process at the time it actually received them. As characters were received they would be packed into one word buffers held in central memory, and a central action would only be required when the one word buffer filled, or when a break character arrived. Thus we attempted to hand the full duplex facility of a TTY to a user process, and still keep the number of interactions low. Except under unusual circumstances, interactions with a central program would occur only once per one word buffer. Thus the number of interactions was reduced by a factor of five (the number of raw teletype characters which could be held in one central memory word). The central program itself transferred the words to and from a buffer in an ECS file, and only interacted with the user process when that buffer was full or empty, another reduction. (Of course, there had to be interactions for break characters etc.)