The raw/cooked stuff isn't the source of the worst hair in the TTY driver; that would be delays (on the output side), and delimiter processing (on the input side). The delays are for mechanical terminals, because they need delays after a motion command (e.g. NL, CR, etc) before the next printing character is sent; differing for different motion control commands, further complexified by the current print head position - a Carriage Return from column 70 taking a lot longer than one from column 10. The driver keeps track of the current column, so it can calculate this! It does the delays by putting in the output queue a chacter with the high bit set, and the delay in the low bits; the output start routine looks for these, and does the delay. On the input side, every time it sees a delimiter (NL, EOF), it inserts a 0xFF byte in the input queue, and increments a counter to keep track of how many it has inserted. I _think_ this is so that any given read call on a 'cooked' terminal will return at most one line of input (although I don't know why they don't just parse the buffer contents at read time - although I guess they need the delimiter count so the read call will wait if there is not a complete line there yet). I should look and see how the MIT TTY driver (which also supported 8-bit input and output) dealt with these... Noel