Four wires, two computers, one historic test

@Mark Ollig

Long before billions of devices connected over the internet, the US military sent digital information over telephone lines.

During the late 1950s, SAGE, short for Semi-Automatic Ground Environment, was an early interactive, networked Cold War-era air-defense system.

SAGE sent digitized radar data over telephone lines to computers using early modems, short for modulator-demodulators, that turned the radar information into analog tones the telephone circuits could carry.

Those computers processed radar tracks at SAGE direction centers and helped console operators watching radar screens direct interceptor aircraft and surface-to-air missile batteries toward approaching bombers.

SAGE demonstrated that telephone lines could reliably carry digitized radar data across long distances to computers used to defend the nation’s airspace.

Duluth’s SAGE Direction Center began operations Nov. 15, 1959, adding Minnesota to the northern tier of SAGE sites watching for potential Soviet bomber attacks.

SAGE had already shown that telephone lines could carry radar-derived information as analog tones to air-defense computers, but that was not the same as using telephone lines for direct digital data exchange between computers sharing programs, information, and processing work.

Lawrence G. “Larry” Roberts and Thomas M. “Tom” Marill took that next step, testing whether distant time-sharing computers could exchange digital data over telephone lines and share computing resources.

J. C. R. Licklider, a psychologist and computer scientist, wrote an Advanced Research Projects Agency memo April 23, 1963, for the “Members and Affiliates of the Intergalactic Computer Network.”

In it, he described how researchers could use connected computers to share programs, files, information, and remote computing power.

Licklider imagined his “Intergalactic Computer Network” choosing the easier path: either sending a user’s information from one connected computer to a program running on another, or bringing the remote program back to work on the user’s information.

In 1965, Roberts and Marill began putting that vision into practice with an experiment linking two distant time-sharing computers, machines that divided processing time among several connected users.

The project connected the TX-2 computer at Massachusetts Institute of Technology’s Lincoln Laboratory in Lexington, MA, with the Q-32 computer at System Development Corp., or SDC, in Santa Monica, CA, about 2,590 miles away.

A limited 1965 Lincoln Laboratory-to-SDC test may have come first, but it was not a full computer-networking session; the two locations only sent and received bits to test whether a cross-country telephone line was reliable.

Because the TX-2 did not receive its modem until mid-1966, Roberts said he probably connected the line through an analog-to-digital converter.

Later 1966-67 records identify the TX-2-to-Q-32 circuit as a Western Union telephone-data circuit, not an ordinary public telephone call.

Western Union’s Broadband Exchange Service, launched Sept. 30, 1964, in Boston, gave businesses an alternative to AT&T’s Bell System facilities for data, facsimile, and voice.

Customers could dial into a broadband connection, select a 2 or 4-kilohertz bandwidth on a toll basis, and use Western Union’s coast-to-coast microwave link to exchange computer data.

The Bismarck Tribune reported Nov. 20, 1964 on Western Union’s $80 million microwave network, about $860 million in this year’s dollars, which supported Western Union’s Broadband Exchange Service.

The 7,500-mile network, designed for about 7,000 voice channels, connected Boston, New York, Washington, DC, San Francisco, and Los Angeles.

Its 267 microwave stations could carry high-speed facsimile, or fax, transmissions; computer data; telegraph messages; voice calls; and Telex, a typed-message service using teleprinter machines.

Western Union supplied modems to convert computer data to analog signals; customers provided equipment and wiring; and users managed calls and data channels with a special push-button Western Union phone.

The service transmitted up to 4,800 words per minute, sent a fax page in under three minutes, and allowed simultaneous sending and receiving.

For the 1965 TX-2-to-Q-32 project, later records identified Western Union Broadband Switching Service, a 1,200-bit-per-second asynchronous data set, and automatic answering equipment at SDC, which answered the incoming data call from the TX-2 side.

AT&T’s Bell System supplied Western Electric Data-Phone data sets, or modems, that linked business computers and data terminals to Bell telephone lines.

A data terminal was a device, such as a screen-and-keyboard terminal, teletypewriter, or card reader, used to enter, send, or receive information.

Computers or terminals connected to the data set through a serial interface cable, and the data set converted digital signals into tones for standard telephone lines.

The Bell 103A transmitted up to 300 bits per second over standard telephone lines; the Bell 201A reached 2,000 bits per second over dialed long-distance voice-grade lines.

The Western Union 1,200-bit-per-second data set sat between them: faster than the 103A, but slower than the 201A.

The larger difference was the network: Bell System facilities versus Western Union’s separate, measured-rate microwave broadband network.

By early 1967, a TX-2 user started the AT program, short for Algebraic Translator. The TX-2 automatic dialing equipment printed “dialing SDC,” then “connected.”

The TX-2 and Q-32 exchanged data over the Western Union four-wire, voice-grade path with separate transmit and receive paths; modems converted bits to audio tones and back.

Surviving records do not identify the dialing digits, address code, signaling method, switching office, or internal Western Union routing.

After connecting, AT logged into the Q-32 system, loaded LISP (list processing), sent a LISP program for compilation, and waited.

Roberts and Marill’s “Toward a Cooperative Network of Time-Shared Computers,” published in the 1966 Fall Joint Computer Conference proceedings, became an important early paper on computer resource-sharing networks.

It described the TX-2-to-Q-32 project, showed how two time-sharing systems could exchange data across the country, and explained resource sharing: one research center using another’s resources rather than duplicating programs and machines.

Roberts later said the concept proved computers could work together.

In the Roberts-Marill test, message blocks had to arrive intact; one flipped bit could ruin a number or command. Circuit switching kept the coast-to-coast circuit reserved for the session, even during short data bursts.

In the 1960s, packet switching offered a different approach: It broke data messages into small, addressed pieces called packets.

Those packets could take turns moving through shared network connections, then be put back together at the receiving end.

That made networks more efficient because many users could share the same connections instead of tying up one dedicated circuit for one session.

Packet switching helped shape the Advanced Research Projects Agency Network, or ARPANET, which went online in 1969 and became the foundation of today’s internet.

By the 1980s, the 1960s data services described here were giving way to faster and more flexible options, including end-to-end digital data services, dedicated T1 private lines carrying 1.544 megabits per second, dial-up modems moving from 300 to 1,200 and 2,400 bits per second, and public packet-switched networks.

In the 1990s, Internet Protocol networks pushed that evolution further as fiber-optic transport and higher-capacity digital circuits carried more long-distance data.

By the 2020s, packet-based networks carried webpages, email, cloud applications, streaming video, video meetings, text and chat messages, social media platforms, internet-based voice calls, and cellular voice calls.

Today’s networked computing world all goes back to four wires, two computers, and one historic test.