Minitel would have turned 40 this year

Just in time for the 40th anniversary of its 1978 launch, Minitel: Welcome to the Internet (Mailland and Driscoll; MIT Press) is here to remind us that a consumer Internet was pioneered by the French.

Let me do a few posts on this book. Let’s start with the overall system architecture. We tend to think of TCP/IP as inevitable, but it was not the world’s only way to do packet-switched digital communication. Minitel ran on X.25. Terminals would dial up switches, which would then connect the terminal to the server of the user’s choice.

The authors explain the overall concept:

In the early 1980s, growing a platform like Minitel was especially difficult because the general public did not yet see the point of going online. … One strategy for attracting a critical mass of users and service providers to a new network is to “prime the pump” … This is exactly what happened in France. The State ordered millions of terminals from private manufacturers (which prompted the creation of new manufacturing lines) and gave away the equipment, free of cost, to every French telephone subscriber … The State further incited this fresh user base to actually connect to the network by creating a free, online phone book, l’annuaire électronique. Finally, it substantially lowered the barrier to entry for end users by not implementing an up-front subscription model but rather charging users based on their connection time and adding the resulting fees onto users’ monthly phone bills … To support the creation of new services, the telephone company rebated about two-thirds of the connection fees to the service providers.

By default, Minitel terminals were designed to decode and display pages of data in videotex format. In the late 1970s, videotex referred to an emerging family of media technologies intended to bridge the twin pillars of midcentury telecommunications: the telephone and television. The simplest implementations of videotex broadcast a revolving set of static “pages” including a mix of text and images to be displayed on a home television. One can imagine such a system being used to circulate announcements about weather, civic matters, or a calendar of local events. This was called teletext.

Télétel was the official brand name of the overall ecosystem that connected Minitel terminals to videotex services.

Crucially, it created a gateway between the preexisting, switched telephone network and newly built public data network, Transpac. Transpac was a packet-switched network using the (then cutting-edge) virtual circuit X.25 protocol.

How about the terminal hardware?

The [early 1980s] Minitel 1B is a self-contained data terminal. All the hardware fits into a molded plastic case approximately eight inches tall, ten inches wide, and ten inches deep. The front of the Minitel 1B is dark brown, and the rear of the case is a lighter beige color. The keyboard is attached to the bottom edge of the front of the case by a hinge. When not in use, the keyboard folds up and is held in place by a latch at the top of the case. In the rear, the case bumps out to accommodate the electron guns of the built-in cathode-ray tube. A large knob embedded in the case controls the brightness of the display. On one side of the rear of the case, a small plastic window lifts up to reveal a serial port with a five-pin DIN connection. On the other side, two long black cables lead out. One cable ends with a standard “Type-C Europlug” AC power plug, and the other ends with a standard T-plug for connecting to the French telephone network. Unlatching the keyboard reveals a small glass screen, or l’écran, approximately eight inches across. A square power button and small red LED are below the screen. The keyboard, or le clavier, includes sixty-four keys. The keys are “chiclet” style, similar to the Sinclair ZX-80 or IBM PCjr, with a few millimeters of space between each. In addition to twenty-six letters, ten numerals, nine punctuation keys, and a space bar, the keyboard includes both a shift key and unlabeled modifier key. The effect of the shift and modifier keys is indicated by a set of color-coded alternate characters printed on the surface of the keyboard. Ten keys at the top of the keyboard correspond to Télétel commands, such as sommaire (table of contents) and envoi (send).

A true “dumb terminal,” the Minitel supplied a reliable interface to remote information services while performing minimal computing of its own. It is often said that the Minitel did not have a CPU or memory, but this is not strictly true. While the Minitel 1 was not a generally programmable computer, a microprocessor was necessary to operate the modem, serial port, keyboard, and screen. Similarly, the Minitel 1 did not have an operating system or BIOS but rather a small set of software burned into ROM that implemented decoders for the keyboard and display as well as protocols for serial communication and error correction. Additional volatile memory was used to store the current state of the screen and provide a buffer for user input. More expensive models such as the Minitel 10 included additional hardware and software features, including a built-in tone generator and automatic telephone dialer.

The Minitel modem was an asymmetrical duplex modem—a communication standard that had been continuously negotiated and refined by the International Telecommunication Union (ITU) since 1964.27 In practice, this meant that the Minitel could receive (demodulate) data at 1,200 baud while simultaneously sending (modulating) at 75 baud.

One unusual feature of the Minitel 1B was the option [when uploading a large file] to “flip” the transmission rates (retournement du modem).

Although the official documentation only hinted at this use, the Minitel could act as an external modem for a standard PC. This was an enormous opportunity for microcomputer hobbyists in France. Whereas a comparable PC modem might cost fifteen hundred francs, the Minitel was free. The process of connecting the Minitel to a PC was not simple—both soldering and programming were required—but the cost-saving hack was documented in widely circulated magazines for the microcomputer enthusiast.

An obvious commercial application?

Smart cards were already in wide use in France as bank cards and prepaid phone cards. The prototypical Minitel card reader was the LECAM, produced around 1987 and rented to Minitel users by France Telecom for a monthly fee. The reader was the same width and depth as the Minitel 1B, and designed to be clipped to the top of the terminal. The reader contained a generally programmable computer that could run software stored on smart cards. This promised to extend the functionality of the standard Minitel in a variety of ways, including turning it into a point-of-sale terminal for processing credit and debit card payments. This application was especially notable at a time when most US merchants still used manual “zip-zap” carbon imprinters to record card payments.

Careful attention to standards?

The DGT attempted to encourage the creation of new services by publishing high-quality documentation of the Minitel standard as well as pamphlets detailing design patterns for efficient, user-friendly systems. The investment in developing Télétel services was further protected by the stability of the system. The core characteristics of the platform stayed constant for the system’s life. In contrast to the rapidly changing microcomputer market, a Minitel service written in 1985 would continue to function, unchanged, in 2012.

How about the servers?

The typical Minitel server was built on a minicomputer running a multitasking operating system such as Unix. The capacity of a server to handle multiple simultaneous connections was limited, on the one hand, by its network interface and, on the other hand, by its host software. Before Minitel rolled into full-scale production and ran over the public data network, Transpac, servers were connected to the telephone network by a bank of modems, each of which could handle one user at a time.

Unlike the British implementation of videotex where the production of content and its distribution were completely integrated, the CCETT separated the network protocol (which developed into X.25) from the protocol controlling the visual display of data (Antiope). By separating these two layers, the CCETT videotex system was effectively “medium free.” In other words, the French standard for videotex/teletext did not proscribe one or another medium for transmission.

The points d’accès vidéotex, or PAVIs, were specialized computers sitting in the logical center of the platform responsible for overseeing the exchange of data between old and new communication media.

To provide a gateway function, the PAVI was composed of five interfaces: a user-facing directory service, a database of known Télétel services, a database of known subscribers, an RTC modem, and an X.25 Transpac modem.

Each PAVI was built on an Alcatel-CIT E-10 switch. Whereas the Minitel 1 is a small, approachable device, the E-10 is massive, filling six large cabinets in a temperature-controlled environment.

Why did the French push this out to consumers? The authors say that deficiencies in the state-directed telephone system provided motivation:

At the end of the 1960s, France had one of the worst telephone networks in the industrialized world. The waiting list for a copper pair installation for 90 percent of clients was three years while at the same time in the United States, 99 percent of installs were completed within three days. In 1971, the penetration rate in France was equivalent to that reached by Denmark in 1930, Sweden in 1935, the United Kingdom in 1956, and Italy in 1964.

In 1975, the government decided to overhaul the country’s patchwork telephone network and replace it with a completely automated system. Consistent with the ambition of previous grands projets, the new plan—dubbed “A Phone for Everyone” (un Téléphone pour tous) explicitly sought to provide universal service to both voice and data. For data, a high-speed public packet-switched network was necessary. An upgrade of this magnitude was expensive, however, and the fees generated through telephone calls alone would not be sufficient to cover the added cost. Instead, it was necessary to not only improve on the existing network but also develop novel revenue-generating services that could recoup the heavy investment.

The French also wanted to build their own thing so as not to be dependent on IBM, which they saw as likely to dominate the world of computer networks as it had dominated mainframes(!).

How successful was it? There were 800,000 terminals and 500,000 active users… at the time of the system’s shutdown in 2012. At the peak (early 1990s), roughly one fifth of French telephone subscribers had a Minitel terminal. There were over 6 million terminals and nearly 90 million connection hours per year (Minitel was expensive to use, so people limited their time online).

More: Read Minitel: Welcome to the Internet