eas-lab@absamail.co.za writes:

]The char-string "ATA" from the local serial port is unambigious, except
]for the speed. Hence logically, at this stage the modem must know
]what speed it's ser-port is talking.
]So the modem must first determine the ser-port's Tx speed ?

For older modems because it only operates at one speed. In newer modems
because the modem detects the speed that the serial line is operating at
(looking at the stop bits, etc) and fitting its own speed to that of the
serial like.

] How ?

Do you really want to know the details of all the gates, etc, on the
modem that accomplishes this?

]> ppp requires 8 bit .
]> You must set up your modem to do 8 bit communication.
]OK, so changing between 7 and 8 bit can't be used to
]distinguish the 2 modes: command and active ?


]How does the local machine know that its modem has "trained with a
]modem on the far side", so that it can send the 7 char "CONNECT"
]string (plus ) ?

Because that is what training is all about. Signals are sent down the
line to the far end, by both modems. When they find a speed at which the
errors dissapear, that is the active speed. If during use the errors
start to become too numerous, they retrain.
They must both settle on a speed and also on an echo cancellation.
Again I do not think you want to know the details, and if you do, I
cannot supply them.

]Apparently the 'negotiated' speed between the 2 modems, does not
]affect the ser-port speed speed(s) ?
]This is handled by the modem(s) data buffers and flow-control ?


] The complexity of ppp is exposed via the FSM state diagrams or
] tables, but the 'modem-automata' seems to be glossed over
] in the literature.

Wrong literature. Get some modem literature. PPP is in a layer above the
modem layer. ppp can run over modems, over serial lines and even over
ethernet. ppp does not care. That lower hardware layer must just deliver
certain stuff.