The Growth of Packet Switching Networks
During the 1970s, inspired by ARPANET, packet switching networks emerged throughout the industrialised world. Sooner or later each would provide the international standard interface, X.25, for access to their networks. X.25 will be described in more detail in the next chapter. In this chapter we provide a summary of some of the networks which used it.
In 1973, the Federal Communications Commission (FCC) granted a license, to offer value added network (VAN) services (over leased lines), to Telenet, a subsidiary of Bolt Beranek and Newman, the subcontractor who had built the switches for ARPANET. In the same year, Laurence (Larry) Roberts, chief scientist on the ARPANET project, joined Telenet as President. In 1975, Telenet opened for business, becoming the first public packet network service (Roberts 1978). Roberts says that by 1978 Telenet was providing services to 180 hosts in 156 cities across the US and that there were interconnections to 14 other countries. In 1979, the Telenet network, acquired by GTE that year, was used to pilot the use of merchant dial terminals, akin to what we in the UK nowadays call PDQ terminals, for VISA (Stearns 2011, 150).
Like ARPANET, upon which it was based, Telenet was implemented as a datagram network internally but had a virtual circuit host interface, unlike ARPANET where the virtual circuit interface was partially implemented in the host (NCP). That meant that the Telenet network was completely in control of sequencing, flow control and error correction. In Roberts' opinion, the alternative was "not saleable as a public service" because of the burden imposed on the customer. In due course Telenet replaced their interface with the emerging international standard, X.25, to whose development they had also contributed. Their solution was a fairly common one at this stage.
Hot on the heels of Telenet came four experimental packet switching networks: Réseau à Commutation par Paquets (RCP) from the French PTT (1975), Datapac from Bell Canada (1976), Experimental Packet Switching Service (EPSS) from the British Post Office (1976) and CYCLADES/CIGALE from the French Ministry of Industry (1976).
Datapac became operational in 1977. In 1974, the decision had been made to make it a datagram network with a virtual circuit interface provided by X.25. Datapac was the first public network to fully support the X.25 family of standards. The Datapac network served for over 30 years and at its peak supported 50,000 devices. The dominant application in its final year was for point-of-sale devices (Rybczynski 2009). In 1979 Datapac was connected to Telenet using the X.75 protocol, a slightly modified version of X.25.
The EPSS experiment was started in 1977 with around forty customers participating. The majority was split, almost evenly, between universities, government laboratories and information technology (IT) companies of various kinds, the remainder being non-IT commercial organisations (Burgess and Lockwood 1976). EPSS had been designed before the availability of the CCITT standard and defined its own protocols. It was followed, in 1978, by the International Packet Switched Service (IPSS), introduced even before a national service, in response to customer demand for access to US databases. IPSS was based on CCITT standards and opened, following a trial service in 1977, in conjunction with US companies Tymnet and Western Union, and later Telenet (Kelly 1981). The national service, Packet-Switched Service (PSS) opened in 1981.
In 1979, IPSS was connected to EURONET, the network of the (then) 9 nations of the European Economic Community (EEC) and, in 1980, to the Canadian networks, Datapac and Infoswitch (Kelly 1981). Post Office/British Telecom (BT) annual reports record the growth of countries connected to IPSS rising from 20 in 1983, with one million calls processed, to 80 countries in 1986. In 1983, a BT staff magazine reported that PSS had more than proved its worth - "High street shops, where is being used to speed up shopping for customers using credit cards, are some of its latest customers" and other uses included "electronic funds transfer, of which cash machines outside banks are an example" (BT 1983). By 1989, a BT marketing document proclaimed that PSS "is now used by more than 100,000 people in organisations of every size and description" and it "presently handles in excess of 1 million calls a day" (BT 1989).
The two French networks took almost diametrically opposed approaches. Louis Pouzin, of Institut National de Recherche en Informatique et en Automatique (INRIA), led the CYCLADES/CIGALE team. Like ARPANET's NCP, the transfer station (ST - station de transport) of CYCLADES dealt with host-host interactions. Pouzin's idea was for the ST to present a common user interface to heterogeneous hosts and from there "the design proceeded inwards down to the packet switching interface" (Pouzin 1975). The design was intended to "facilitate interconnection with other networks" and this involved "putting as few features as possible at levels buried in the sensitive parts of the network". Accordingly the transit sub-network, CIGALE, would not fragment messages but would accept only packets. It would take no responsibility for sequencing nor would it attempt to impose flow control. By 1974, the network had 16 hosts, in universities and research establishments, connected to five nodes, in cities the length and breadth of France.
Like the work of Baran and Davies, Pouzin's was an idea whose time had not yet come. He recognised that his approach was radical and that the difficulty lay in finding a solution which could and would be effectively adopted across a wide variety of hosts. Where Pouzin's design was pure datagram, Rémi Deprés's RCP network was virtual circuit, in implementation as well as appearance. The appeal of the approach for him was that it made it easier to exercise control over quality of service, by only allowing those circuits to be established for which there are adequate resources, and it made better use of the lines by carrying less addressing information (Després 2010). The latter was particularly important at the time given the poor quality and low speed of (analogue) lines.
RCP was successfully implemented with 3 switches, in Paris, Lyons and Rennes in 1975. Transpac, RCP's commercial successor, went live in 1978 implementing the X.25 virtual circuit interface. By 1980,
with 2,395 operational X.25 access links, banks counted for 28% of established VCs [virtual circuits], service bureaus for 19%, industry for 15%, and administrations [sic] for 14%. One year later, more than 5,000 X.25 accesses were operational (at a time when Internet was still in its infancy, with its first 213 hosts) (Després 2010).
The first transatlantic links were opened in 1979 closely followed by links to European X.25 networks in Germany, the UK, Spain, the Netherlands and Luxembourg as well as with EURONET.
Some private packet switched networks of the period are also worthy of mention. The US Federal reserve became early adopters of X.25 in 1980 to accommodate its internal data communications requirements (J. Martin 1984) and the Société Internationale de Télécommunications Aéronautiques, now known simply as SITA, developed a packet switched network to provide services to the international air transport industry (Chretien, Konig and Rech 1975). In the 1980s the SITA network supported 120,000 terminals and printers, served 460 airlines across 31,000 offices in 187 countries and processed 36 billion data transactions and messages each year (Rybczynski 2009).
It is clear from the few examples presented here that the early packet switching networks were instrumental in stimulating another step in the globalisation of certain business sectors. In doing so, they were helped by the X.25 standard, and its progeny X.75, which made it possible for different networks to present a single face to their users worldwide and also to interconnect with each other.
1. An unfair comparison given that ARPANET was still confined to hosts in government locations and DoD sponsored university departments.
Pouzin, Louis. "Presentation and Major Design Aspects of the CYCLADES Computer Network." Proceedings of the NATO Advanced Study Institute on Computer Communication Networks, Sussex, 1973. Leiden: Noordhoff, 1975. 415-434.