| • Science | • People | • Locations | • Timeline |
Prior to the introduction of the Bulletin board system, modems were typically used on dedicated point-to-point telephone lines that were always connected to a modem at the far end. In the situations where this was not the case, the user would dial the phone manually before connecting, or pick it up if it rang. In a few cases the computers themselves were required to call a selection of numbers, and for this task a separate "dialer" was used, plugged into a different input/output port on the computer (typically an RS-232).
When Dennis Hayes was creating his first "smart modem" in 1977, he faced a problem that no previous modem manufacturer had: how to tell the modem what phone number to dial, using only one port. One solution would be to use a separate set of "command pins" that would be dedicated to sending and receiving commands. Another would be to use a signal pin that indicated that data on the main data lines should be interpreted as a command.
Instead, in order to save money, Hayes chose to re-use the existing data pins with no modification. The modem could be placed in one of two modes, data mode in which the modem sent the data to the remote modem (when the modem is in the data mode, it assumes that everything it receives from the computer is data and sends it across the phone line), and command mode in which the data was interpreted as commands to the local modem (a command to execute by the local modem). To switch from data to command mode an escape sequence string of three plus signs followed by a pause of about a second was sent (to go to the command mode without losing the connection), and to switch back you sent the online command, O. In actual use many of the commands automatically switched to the online mode after completion, and it is rare for a user to use the online command explicitly.
This use of in-band signalling leads to a potential serious problem: what if the data being sent in data mode contains three plus signs? This could happen randomly quite easily, and it would happen any time someone attempted to send data describing the system – this document for instance. In order to reduce the chance of this problem the pause at the end of the escape sequence was required, if any other data was received within one second of the three plus signs, it was not the escape sequence and would be sent as-is.
The command set includes commands for various phone-line manipulations, dialing and hanging up for instance. It also includes various controls to setup the modem, including a set of register commands which allowed the user to directly set the various memory locations in the original Hayes modem. The command set was copied largely verbatim, including the meaning of the registers, by almost all early 300 baud modem manufacturers, of which there were quite a few.
The expansion to 1200 and 2400 baud required the addition of a small set of new commands, some of them prefixed with a & to denote those dedicated to new functionality. However Hayes was slow to move to higher speeds or the use of compression, and three other companies led the way here – Microcomm , U.S. Robotics and Telebit. All three used their own additional command sets instead of waiting for Hayes to lead the way.
Soon there was an explosion of new models, including new ones from Hayes, following one sort of standard or another. Things became simpler again during the widespread introduction of 14.4 and 28.8 kbaud modems in the early 1990s. Slowly a set of command based heavily on the original Hayes extended set using & commands became popular, and then universal. The only other command set still popular is the US Robotics set, due to their widespread popularity.
The command set was unprotected, but the concept of a "guard time" after the escape sequence was patented. In the late 1980s Hayes started enforcing the patent, charging $1 per modem that used it.