B^EMEm s: (UBSAKEESI or rt^^ g SETTING S TANDARDS FOR COMPUTER COmUNICATION THE CASE OF X.25 Marvin Laurence A. E. Sirbu Zwimpfer September 1984 CISR WP #117 Sloan WP ^/s^^^'g^ _DEWEy LIBRARY Center for Information Systems Research Massachusetts Institute of Sloan School of Technology Management 77 Massachusetts Avenue Cambridge, Massachusetts, 02139 S TANDARDS SETTING FOR COMPUTER COmUNICATIQN : THE CASE OF X.25 Marvin Laurence A. E. Sirbu Zwimpfer September 1984 CISR WP #117 Sloan WP -^/^"^/^y-g^ © n, A. Sirbu. L. E. Zwimpfer 1984 Center for Information Systems Research Sloan School of Management Massachusetts Institute of Technology standards Setting for Computer Communication: The Case of X.25 Marvin A. Laurence and Zwimpfer Sirbu E. Introduction 1 The to first international meeting to set standards for data discuss interconnection of to computers and communications technology economic environment of in in of the 2. of particular. the 1970s, coupled with in 1865, these changes, and illustrate The convergence changes these industries have forced dramatic changes standards, and the processes by which standards are developed, broad directions Paris coordinate the development of standards for general and data communications in in telegraph systems. Since that time, an elaborate network of institutions and procedures have grown up which attempt telecommunications communications was held in this in them with a detailed examination the political and both the role of in paper, of we shall outline the of the development packet switching standard, X.25. The Nature and Function Standards are of many types: of Standards the IBM Personal standard for personal computers. Bell Computer Technical Reference defines a de Communications Research procurement specifications represent another type of standards document. More typically, however, v/e think of formal documents facto certified by a self-defined when we think of standards standards making body. Groups engaged in formal standards setting include international agencies, national standards bodies, professional associations, trade associations, government regulatory agencies and cabinet departments and bureaus. Standards may, but usually do not, have the force of law; generally they represent voluntary agreements among producers. Massachusetts Institute of New Zealand Post Office Technology While a Firms can be tliouyht staiKJcird may of as a documenl. compatible devices though v^hat build it is also involves a set of behaviors and a process. required for compatibility down. Conversely, because a standard has been reduced agreed to abide by Standardization it. the patterns of behavior which will is is nowhere does not mean to writing the process of developing the consensus written that firms among have firms and achieve the ultimate goals of a standard, not just the production of a document. the area of computers In and communications there are two compatibility and variety reduction V Compatibilitv or interchanqeabilitv standards are necessary for complementary products, manufactured by and boxcar axles, equipment (DCE) light will Variety reduction work together. Railroad gauges not work together unless they are compatible. number achieve potential economies of scale. induce chipmakers different producers, to bulbs and sockets, data terminal equipment (DTE) and data circuit-terminating limiting the , principal functions of standards: to invest in of versions of a product, A local area VLSI implementations may be essential for producers to network (LAN) standard may be necessary of a local to network protocol. Without a standard, they would be skeptical of achieving sufficient production volumes to justify the front end investment in chip design. Variety reduction can also lead to The absence reduced inventory holdings^. software dealers must stock programs in of a economies of stocking and distribution through standard for 5 1/4 inch floppy disks means that multiple versions for each possible system. While the compatibility and variety reduction functions are often closely related, they are not the same. Maximum economies of compatibility would require a single local area network protocol; but available scale in chip manufacturing two different protocols. This Mirlan In LANs with the same is may be achievable where precisely what Intel proposes to do the in same chip can be used for supporting both Ethernet and chip set^. the information and communications arenas, compatibility has two aspects: 1. the functions or services provided by the tv/o devices must be compatible; and 2. the procedures used to provide the functions or services must be compatible. procedures can often be made compatible Identical functions provided using different simple softwate conversion the functions provided are different, an exact translation If means Additional functionality usually possible. tiigfier research and development. Functionality may also be end As a consequence, arguments over user. costs, either for the critical to tiiroucjh a may not be end equipment or for achieving market acceptance by the v;hat functionality to include in a standard are usually the most contentious. 3. Economic Theory The economic motivation •- Standards of standards for compatibility or variety reduction - be different depending upon the goal will and depending upon the economic positions of the standard of the participants the standards process. in In the absence of economies of scale in production or of inventory, there variety reduction standards. Similarly, there will be a demand for compatibility is no incentive to seek standards only wish to communicate or share information with a wide range of other users, and translation, or multiple basic The market structure standardization. equipment of sets, Users often pay dearly users the costs of significant. buyers and both sellers is absence for the clearly relevant to the incentives of standards, thus, they should for be the Buyers may be fragmented, however, and unable supporters of standardization. principal is if if to coordinate their demands. Where buyers are few and powerful, they can more easily exert pressure on sellers to Concentration standardize, increased Where with, demand standardize in force among buyers should Communications communicate or networks the for have standards by coordinating significant is externalities: it total size of the may be the more other people one can Where these the network. equipment which can access a larger sellers are fragmented, purchase specifications. thus favor standardization. more valuable order to increase the their externalities translate into portfolio, sellers have an incentive to market. difficult to coordinate the development of standards. Where there is one dommcmt firdi and many small ones. v;e can expect de facto standards be set by to the dominant firm (absent patent, trade secret or copyright protection which inhibits copying the dominant Where firm)'*. sellers are few but of equal size, compatibility standards would act to reduce market power, and thus may be resisted. Integrated sellers v;ho produce both primary and secondary products may resist standardization since allows entry to competitors it who produce one only of two conjoint products. The coalescence of firms around a standard positive externalities of adopting the standard of other users. Thus, smaller firms are led to is a positive feedback process which results from the which allows communications with the largest number fall in behind the standard of v/hichever firm or group of firms establishes the largest early following. Standards Setting Processes 4. in Computers and Telecommunications the mid 1970s, the structures of the telecommunications and computer industries differed greatly. In These differences included the characteristics employees. of All of the legal charter of the participants; their technology; and the characteristics their of these differences contributed to quite different patterns economic motivations; the technical of and behavior managerial v/ith respect to standards. 4.1 . Telecommunications Telecommunication services have of traditionally Posts Telegraph and Telephone (PTTs) -•, been provided by government agencies -- Ministries by nationalized firms such as Nippon Telephone and Telegraph, or by regulated private monopolies such as AT&T. Until quite recently, the focus firms has been on the extension of basic population. Thus, from 1976 to 1983, the 24 million. In telephone service number of telephones most countries there was an unmet backlog of to in of these an ever greater fraction of the France increased from 7 demand for basic service; million to consequently, telephone authorities have been more concerned with increasing supply than designing and marketing ne-.v products or services. Technological change v.-as confined primarily to the process of providing telecommunication services and not to the service product. Telecomiminication operations are characterized by large capital investments in highly reliable, k ' long-lived equipment. years, Telecommunications equipment and thus new technologies are introduced equipment continue Because to work with the most network operators, would for in fact spend all working promotion and therefore middle As monopoly providers own standards International of life the technical and with the for its levels organization. Seniority managers are is to no voting power. of has been low; most an important element typically over 40. At the international level, they meet together as members their of the decide on interconnection arrangements between to the telecommunication carriers also Technical standards (or Recommendations as they are set by the International Consultative (CCITT), a permanent organ of the ITU. The work same management The manufacturers who supply equipment known) are an absolute requirement that new telecommunication services, most have absolute authority to set Telecommunication Union (ITU) participate but they have is telecommunication systems and the monopoly position level technical at the national level. national networks. formally of staff mobility at their There slowly. old. character of the specialist often depreciated over a period of 20-40 is Committee CCITT meets every specialized study groups and approve for Telegraph and Telephone four years to agree on a new Recommendations. program Detailed technical work of is carried out by Working Parties or Special Rapporteurs. Telecommunication standardization processes have decades defined of first-hand and homogeneous experience understood tight-knit by in been very international agreement-making. participants^. The participants group with a rather conservative outlook. international standards to motivating factor the in traditionally permit the interconnection of national stable, based on many The process has been have also been a well- relatively The absolute necessity for networks has been a strong obtaining complete consensus on the standards developed, although there have been cases when some participants have placed more value on national goals than international ones, (e.g. color television standards®). Typically, standards have only been approved after they have been implemented and tested; that is, the standardization process has relied heavily on compromises Hov;ever. the process has not been able to cope based on established de facto standards' situations m which participants already have a substantial commitment standard: for example, the divergent U.S. and European standards for 4.2. to v^ell with a particular de facto PCM. The Computer Industry The computer industry in the mid 1970s presented a completely different picture. Instead of stable, conservative, national monopolies, the computer industry w/as a dynamic, rapidly changing industry dominated by a few large multinationals. changes The industry has been highly innovative and product mix. The industry's dominant its in in IBM rose to its position through its Unlike the telecommunication industry, the infrastructure required to support a computer system The people involved are also typically (20-40 age group) and as a group have a because they possess more marketable Attempts at standards setting in much younger much encouraged to than in the telecommunication industry higher mobility --partly because of their age but also the computer industry have been diffused and largely ineffectual. Institute (ANSI) been coordinated by a single organization in do the actual standards development work, including trade associations and Electronic Engineers (IEEE)). (e.g. The the United States) but other organizations are United States, the Electronic Industries Association (EIA)), and professional groups of Electrical is relatively skills. At the national level, standards have generally American National Standards life of microelectronic miniaturization and the corresponding rapid decline costs has created a pressure on users to regularly upgrade their equipment. small. emphasis Most computer equipment has had a design on marketing, rather than process technology. 5-10 years; the rapid advances firm, rapidly Large computer users (e.g. (e.g. the Federal (e.g. in the the Institute Government in the United States acting through the National Bureau of Standards) have also taken an active interest in the development of standards, and where necessary, have adopted their ov;n standards. Because the computer industry has been so dominated by IBM, consensus standards-setting procedures have been (typically the relatively unsuccessful. Only when there has been pressure from a large user government) has progress been made.'' Because IBM has been a full line producer of both basic aiul ccniplomentary products, it lias had little incentive to support standards wtiicii would piecemeal fashion. have enabled smaller competitors to attack early years operated primarily as standalone devices there it in Also, because computers in the loss incentive for standardization. vi^as Users could achieve internal standardization by buying from a single vendor; standardization on a larger scale brought few additional advantages. Rapid technological For example, at standardization. in 1961 IBM was promoting interchange code, but only four years later had changed 4.3. Changing Industry Structure and Ivlajor changes to its 6 change also undermined bit support an code (BCD) as a US standard 8-bit code (EBCDIC). Effect on Standardization Its both technology and regulation altered dramatically the industry structure in efforts in both These changes affected the motivations computers and communications during the 1970s. for achieving data communication standards, and had a profound impact on the development of X.25. Beginning with the Carterphone decision in 1968 and the Specialized Common 1971, the U.S. began opening the telecommunications industry to competition. literally hundreds of companies would be including computer terminals and -• services, including dominance of the U.S.' anti-trust suit Within a few years customer premises equipment - while others would promise a host of This influx of new new entrants challenged AT&T's filed also in was witnessing 1959, minicomputer market, IBM had a period of increased competition. networks of of large hundred and because lost of its failure to computer or utilities even thousands technology. in threat of foresee the importance of the share to a new group of companies such as possible by the spectacular successes standards: traditional Under the Digital Corporation, Data General, and Prime. With continued growth, the industry also began to from the model netv/ork telecommunication standards process. The computer industry an modems digital facilities. all offering a dazzling variety of Carrier decision in Equipment shift away based on hand crafted mainframes towards interconnected of mass produced machines. microelectronics, This transformation, made even more made imperative the need for not only for compatibility, but to achieve the economies of scale offered by VLSI 8 VLSI allows extremely complex devices to be manufactuied tecliiiulocjy challenge is a design find to needed in change. Complex software - such as characteristics: based standards are easier to for in silicon, costs. investments. however, standards are more computer networking and low reproduction large design costs very low unit cost; the sufficient quantity to justify the frdnt-end Standards help create and define such needs. Once etched difficult to at -- exhibits similar economic Unlike hardware, however, software modify and evolve, posing problems in the stability of standards once they are codified. These changes The shift technology and regulation have affected the standards process in from standards primarily for reasons altered the motivations of the participants. of compatibility The need to define as opposed to interconnect two national markets leads a standard in several ways. for variety reduction order to create a market, to: accelerate the standards process for commercial reasons; 1. pressures 2. definition of standards in parallel with research to towards standards in and development on the underlying technology; 3. increased risks 4. the absence of of problems in the resultant standards market experience making difficult the appropriate level of functionality to incorporate Finally, the to hasty development; and the resolution of controversies over in the standard. deregulation and the convergence of computers and communications has vastly increased number standard due The of players. will be agreed to, larger set of or b) some needs to significant be satisfied increases the risk- that either: a) group of users will find the resulting no standard unacceptable. Thus, as a result of deregulation in production and increased competition came under heavy pressure just as the telecommunications industry and a shift towards in computers, the standards processes packet netv^orks were being developed. in mass these industries 5. Computer Communications and Packet Switching Early computer systetns were quite independent was industry not constrained by the with remote terminals were still absence of of each other and therefore the development standards. Even early timesharing computer systems relatively self-contained: typically both the central remote terminals would be supplied by the same manufacturer manufacturers would copy the de facto standard were required to interface need critical for communications line of was at the principal supplier). However, between interface the which linked remote terminals to the modems which converted digital e.g. some standards modems. The computer terminal the computer. "foreign attachments" to the network, the telephone administrations standards for the processor and the equipment supplied by other (or computers with the communications network, standards of the Through were in single and their control the over a position to determine bitstreams into analog signals suitable for the voice network. But voice telephone circuits are costly to the data communications user and totally inadequate for high speed, error-free data communications. During the late 1960s, the concept of packet switching emerged as a "bursty" solution to the problem of providing efficient, error-free service to computers sending Packet sv^itching systems transmit data by means traffic. which include both data and of binary digits call of addressed packets, control information. available for use by packets being transferred between to create, process, and switch packets communications channels. Thus, more efficient it became was declining at a much of computer processing faster rate than that of cost effective to use computer-based switching to make use of the channel. There are two distinctly different ways to' organize the provision of packet switching service: connection-oriented or connectionless, often referred to as Datagram address The channel then different data terminal equipment. Packet switching became economically viable largely because the cost needed groups These packets occupy the transmission channel for the duration of the transmission of the packet only. becomes i.e. service. In virtual circuits vs datagrams. a "connectionless" datagram network, each packet carries the of the destination and is full routed independently from the origin to destination 10 terminal. As a be corrupted is packets may arrive out in transit by errors, or even of seciuence lost altogether. Others may at the dci^tinalion. In a connectionless network, it the responsibility of the endpolnts to resequence packets or request retransmission of packets • result, Virtual damaged in transit. service circuit is the logical equivalent of a physical circuit, or connection, between the source terminal and destination terminal. An originating terminal starts by sending a call origination packet which specifies the destination to which subsequent data packets are to be sent. The call is assigned a short logical call number which can be used to specify where to send subsequent packets. The network takes responsibility for insuring that packets arrive in the correct sequence for requesting retransmission of lost packets. packet takes down shall see later, debates over At the As with leased the virtual circuit. customers can be assigned "permanent As we as they would on a "circuit" • and virtual circuits" virtual circuit end of a virtual call, lines in the a clearing telephone network, which are always in place. versus datagram service played an important role the development of packet netv,/ork standards.* in 5.1 . Standards The exchange for Computer Networks of data over a computer network requires agreement at many levels, physical interconection of the computer to the communication line to agreements on the the bits exchanged between computers at opposite ends of the network. Early packet networks it was recognized that a standard terminal should connect to the netv;ork. would be needed This interface, between boundary between equipment provided by the carrier in to define from the meaning of the development of how a computer or DTEs and DCEs, would mark the and equipment provided by the user, or the computer manufacturer. The development of standards process and motivations for this interface for standardization in the provides a dramatic illustration of the differences computer and the telecommunications • For a further discussion of the merits of virtual circuits versus datagrams see 8 9 10 11 ' * industries. in 11 A Case Study 6. The revolution in X.25 of computers and coinnuinications has been so rapid and computer communication networks are today so widespread that demonstrated the possibilities of seems it diflicuit to imagine that the packet switching a mere 15 years ago. Indeed, v/hat ARPANET is striking first about the history of X.25 are the doubts which prevailed throughout the process as to the very need for public packet networks, and tiie possibilities of developing a successful standard. While today X.25 can be viewed as a successful standard, there were, as we 6.1. many rough spots along the way. X.25 gets on the standards agenda The need CCITT systems and standards for special with the creation of a Working Party Transmission Working Party was elevated had produced a set of in to for data communications was full study group status, and by 1964, the to had outlined a program should be standardized of to permit international the discussion within to investigate the possibilities of Vli (New Networks 23 points under the general Question: NRD prior to new for Data Transmission) and "What general characteristics data communications over public data networks?" 1972 had concerned between 1968 and 1972 several PTTs had announced circuit switched public data networks; their intention to build such networks.^** Only 23 points. Question C, dealt with packet switching: "Should the packet-mode of operation be provided on public data netv;orks, and if so how should Special Rapporteur had been appointed for question NRD new study group data transmission services.'^ By 1972, the Working Party had became Study Group of the recognized by the the late 1950s. At the 1960 Plenary Assembly, the Data Working Party' was created networks that might be wholly dedicated fvlost of first recommendations dealing with data transmission over the telephone network. At the 1968 Plenary a Joint one shall see, Nouveau Reseau de Donnees C and it be implemented?"^^ By July a program of study established. of 1974 a 12 switched networks 6.2. Early packet By 1972, when the CCITT was still asking whether there should even be public packet switching networks, there were numerous private or experimental net-.vorks already the U.S.'^. networks, NPL RETD Spain and the Tymnet network in the U.S.^'' were also Most telephone administrations, however, expected demand voice and data, to meet the for circuit of traffic, and data was In •• and the idea as it still is -• of skeptics into believers.'^ Hov;ever, there wisdom be networks, carrying both more economical than far communications generated by demonstrated far for carrying the greater both voice still the first International manufacturers v;ould be reluctant to make Conference on Roberts, then head of DARPA, converted many a great deal of uncertainty among PTTs concerning to Larry was at packet services: standards v/ould take too long of establishing public viewed as speculative. Computer Communications, and, according the digital to in operation. in packet networks as the dominant technology ARPANET was October 1972, the switched data services and, ultimately, building a separate packet network just for data. Voice amount Two the UK, and the international airline reservation network, SITA.^^ in in ARPANET operation; in to develop and computer the necessary changes to their products to support a standard network interface. Such skepticism had not deterred ARPANET builders, Bolt first Packet Communications Inc and then Telenet Beranek and Newman) from seeking authority in US the (a subsidiary of to establish public packet networks. By April 1974, both had received approval, though PCI soon dropped out when had trouble raising In capital. France, a decision to build a public packet switched network November (TRANSPAC) was made in RCP.^ In 1973,^® based on experience gained with an experimental network called Canada, a trans-continental Communications Group service it was recognized all of the to packet switched service, digital private line service was introduced in 1973 by the Computer Trans-Canada Telephone System (TCTS).^^ However, be uneconomic for low volume users. DATAPAC, which was announced in TCTS began planning October, 1974.^^ In private line in 1973 for a the UK, the British 13 Post Office av;arded a contract to Ferranti Ltd. August 1973 in for an experimental packet service. EPSS.^^ There was also considerable interest European Informatics Network EURONET was while (EIN) in two packet networks being developed in at the end 1971 Japan's which was a project of the PTTs, designed to A contract 1973, and tested by NTT NTT and attending Commercial Pressure November circuit for a of 1974, at the third meeting netv/ork service would have to provide to the end of 1974. switched networks. use the network. in Needing a standard particular were anxious into operation until 1975, to Rapporteur's meetings. of the Rapporteur's group, attention finally to the question of a standard for the interface it. custom software and hardware for every moved past between the of packet brand of computer With such a standard, the computer vendors themselves might see a standard defined. While DATAPAC was The Canadians not scheduled to come to bo made ICCC in 1974 suggested that there would be many networks at the second in 1980. in If 1974. Moreover, the growing interest in a standard could not be agreed before the 1976 Plenary, then by 1980, each network v;ould have been firmly committed to unacceptable. DTE/DCE Without such a standard, a would-be vendor by the end of the subsequent CCITT study period of in design decisions had packet switching exhibited and the adoption for the v/ork of the Rapporteur's Group, eventually eventually provide the necessary capability for their products to use packet networks. in and successor, DDX-2, for which research began of the ail network and the computers which would use wanted let at Standard whether there should be packet nets that It's took an active interest subfnitting 15 contributions, In EIN was for information databases an integrated circuit/packet switching system, DDX-1, 1974.^^ in was based on separate packet and interface for DDX-2^^, 6.3. make science of 1975. NTT had begun research on installed Europe: The and EURONET. EIN was driven by the research community.^* throughout Europe more easily accessible. EURONET jointly in a new international standard would be both politically its own design, and economically 14 Dave Norton, Thus, his in Communications Group in role as Assistant TCTS began an interface standard. Morton realized that Vice-President for Planniny Computer the of international lobbying effort to press for the adoption of an order to influence the CCITT, he would in support from other telecommunications administrations. A joint contribution first have to would be more obtain likely to gain acceptance than one submitted by a single administration. Horton began the process by issuing a glossy-covered specification of the proposed 6.4. Virtual Circuits The TCTS there was was to service and circulating it widely. ^^ Versus Datagrams draft called for a standard was implementing a DATAPAC virtual circuit be a standard, it was based on a datagram mode network, and the French clear to Horton that of operation. RCP was agreement on However, Telenet also a virtual circuit design. this If fundamental design choice imperative. Roberts, by now President of Telenet, favored virtual circuit service, largely for economic reasons. Simply routing packets represents only 10% of the value added of a packet network; providing circuit service interface, the and other interface computer processing virtual circuit interface facilities to would be the major contribution. With a datagram resequence packets could be provided by the end would force the customer to obtain -- and pay for •• virtual user. A these services from the carrier.^ Technical arguments for choosing that a virtual circuit service datagram network places the various netv;ork control packet sequencing--in the hands of the end user. As a situations only by throttling all traffic were advanced as well. Roberts argued functions--e.g. error control, flow control, result, the from an entire host, even if network can respond to overload it is only one virtual circuit that is generating the overload condition. Datagrain proponents argued that the overhead involved for all simple point-of-sale, or automatic that will feel is required, teller fvloreover, they argued, in setting up a transactions where one packet no matter how reliable the virtual circuit in was wasted each direction might be virtual circuit service is, they must protect themselves against the possibility of network errors by duplicating users in the 15 end computers much of the error virtual circuit service. Finally, a datagram network than it correctiny anil packet sequencing packet voice suffers less from an occasional does provided as part of a facilities lost or froin the uncertainties in delivery delay damaged packet in associated with virtual circuits.^ Faced with these arguments, and recognizing committed call to their respective implementations, DATAPAC Hov;ever. service. TCTS agreed retained the and the French PTT were more that Telenet to align DATAPAC by providing virtual datagram mode as the underlying transport mechanism."^ 6.5. In Formal and informal processes January of 1975, a counterparts in group from TCTS traveled Table 6-1 London, Paris and Madrid DATAPAC the PTTs, and discuss the informally. A By the end of the year, Telenet total of ten this time, interest in such informal meetings, some and the BPO much was followed by a in their series of CCITT meetings (See the formal of the drafting lasting as long as a were also drav/n packet nets had grown meet with to the French PTT, aimed at developing a Although both groups were also participating a chronology of working group meetings), for This proposal. TCTS and informal meetings during 1975, mainly involving draft standard. to in to work was carried out week were held in 1975. these informal meetings. Also by where Rapporteur's Group meetings were drawing to 40-50 people. The same individuals Subcommittee Meetings of of were also representing the Data the International Standards Organization's TC97, as well as Communications in ANSI X3S37. these groups provided additional settings for discussing X.25. 6.6. National Delegations As a their organizations in treaty organization, it and Standing is States •- the US, the Departments of State and national position for in the CCITT and not companies Commerce CCITT meetings. For all its •- that have official organize Study Groups of standing US in the ITU. In firms to formulate a enthusiasm, Telenet could not speak for the US, and 16 Table 6-1: Chronology of Formal CCITT Meetings Related to X.25 Dote Meeting December 1972 5th CCITT Plenary Assembly (Geneva) January 1974 SG VII First Meeting Rapporteur on Packet Switching First August 1974 Rapporteur on Packet Switching Second Meeting November 1974 Rapporteur on Packet Switching March 1975 Rapporteur on Packet Switching May 1975 Rapporteur on Packet Switching SG VII - Meeting Third Meeting Informal Meeting • Fourth Meeting Second Meeting September 1975 Rapporteur on Packet Switching February 1976 SG VII Final September 1976 6th CCITT Plenary Assembly (Geneva) November 1976 Interim Rapporteur on Packet Switching Fifth Meeting February 1980 SG SG SG SG VII Final November 1980 7th CCITT Plenary Assembly (Geneva) April 1977 April 1978 April 1979 could not, on its VII Second Meeting VII Third Meeting FCC 1974 and 1975 the US TCTS experienced organization for the Meeting its in April official development and approval (CTCA). Meeting VII First own, advance public carrier from the of views directly. 1974, position had no it -- problems in Indeed, right until even to it received approval as a recognized attend CCITT meetings. Throughout dominated by the skeptical views an interface standard similar CCITT Meeting in for of AT&T of a was that packet switching was premature. obtaining support from the official representative Canada, the Canadian Telecommunications Carriers Association TCTS's proposals were opposed by a competitor, CNCP, who considered development •- standard would not be in their that the rapid business interest; consequently, TCTS's views could not be presented as the Canadian position. Thus, while Telenet and TCTS played key roles in drafting the eventual standard, the drafts had to be submitted as national contributions from France and the UK, whose positions were determined solely by their respective PTTs. 17 Ifi 1974 International tlie membership CCITT meetings A •- in in Processing was granted information that representatives of the IFIP Network' an advisory capacity. of a IFIP did datagram quickly discovered that they were quite powerless category (d) Working Group, 6.1, submit a number of written service, but representatives to the the in CCITT forum. (1976) draft of X.25 first CCITT tlie for and strongly supported the inclusion contributions, 6.7. X. 25 CCITT. This meant in tlie could participate Federation was finally prepared by an informal drafting party work which was subsequently confirmed informal meetings during the summer in led to the May 1975 in Ottawa in Ivlarch, Numerous formal Rapporteur's meeting. 8 written contributions to the 1975 September Rapporteur's meeting, the last scheduled meeting of the 1973-76 study period. Although there were a no opportunity be translated meeting of for further for full any more meetings into the three SG 12 months remaining Vil in at the study period after September there would be in the Rapporteur level. The results of the meeting needed working languages used by the CCITT and distributed before the February 1976. After the February change before formal ratification at the the final meeting of the Rapporteur's Group was SG VII of the September 1976 CCITT Plenary Assembly. Thus under crucial: traditional September Rapporteur's meeting, however, there were issues: e.g. packet lengths, packet length indicators, call progress indicator for virtual calls, datagram/virtual datagram call datagram and classification be changed from representatives from France and TCTS. CTCA spoke virtual call flow control issue remained unresolved. CCITT procedures, made it seem still "essential" to in 1976. signals, "more data" mechanisms. "^^ Furthermore, the "additional." to the meeting that the He was supported by However, representatives from NTT, ISO, ECMA, was the numerous unresolved and advisory Roberts had proposed against hasty action; as a result the point Rapporteur's meeting final meeting, there would be no opportunity decisions of this meeting would determine whether or not a standard would be adopted By the end to left open. The final USA and report of the doubtful that a standard would be ready for adoption by SG VII. Immediately following the September 1975 Rapporteur's meeting (which had lasted two days), 18 TCTS. Ihe Fiencli PIT and Larry Roberts spent a proposal. mid-October, In TCTS arid Telenet met A key issue their in discussion was flow control. significance only (DTE-DCE); Roberts supported an Despres also supported an attempts control of the draft to eventually reached by disguising if end October, of Washington. D.C. While local flow control significance (DTE-DTE). while Roberts supported in the Telenet and e.g. some of the RCP window networks. differences of opinion to pay in was mandated but endto-end local flow control desired by individual network providers. Everyone were prepared all thie arrange a further X.25 meeting. scheme would create problems when attempts were made the future, but in At redrafting the X.25 Despres supported flow control with endto-end explicit credit flow control recommendation; was acceptable specification Geneva resolving these differences failed, but with assistance from Anthony Rybczynski of at TCTS a compromise was wording in ISO TC/97/SC6 This conflict arose because of different implementations rotation. Initial of Washington. Roberts seized on the opportunity in week for furtfier discussions. Remi Despres from the French PTT attended a meeting Despres was fiitllier knew that flow such a loose to interconnect different this price in the interests of getting the the networks in agreed portions of X.25 accepted. The BPO an result of this v/ere prepared to support. (BPO) this February they described the draft standard as a in Thus, by the end of 1975, with the years and interface. of its Spain supporting it standard was reflected fait their paper even before the in RETD network had been the development of X.25 in their network. who had to do so was prepared in operation for almost national traffic continued to increase without the benefit of a standard felt SG accompli. network or had made a commitment support the proposed X.25 standard. The Spanish five to a exception of Spain, every administration and carrier either already established a public packet to agreement was tantamount 1976 AFIPS National Computer Conference. Submitting for the meeting That by Rybczynski (TCTS), Wessler (Telenet), Despres (French PTT) and Wedlake article drafted VI! meeting was a draft recommendation that TCTS, Telenet, the French PTT and the DTE/DCE was an "unstable study" and therefore had no intention 19 Despite acceptance by the leading earners, the draft SG from 20 countries represented in probleins surfaced. The US official still needed to be ratified many During debate on the draft proposal VII. delegation was a particularly vocal critic: the US of the position no standard should be accepted which had obvious technical flaws. The technical points were not resolved, but a document was agreed US to after the French cooperation on unrelated issues the eleven recommendations of SG at the Plenary. -- VII In was understood September 1976 X.25 including known was that of conflict delegation, under pressure from France, agreed to stop challenging the proposal. Consent by the U.S. for members by the 200 at the to be CCITT in return Plenary, were adopted unanimously, thus -• completing the formal process. 6.8. Problems and Revisions Some saw Adoption of X.25( 1 976) by the CCITT did not bring about immediate widespread support. it as little more than an "advertisement for were many ambiguities and unresolved alignment of the link level of a standard."** The participants themselves Tv/o issues. X.25 with the HDLC of these were of knew there particular importance: standard being developed by ISO; and datagram service. X.25 had been developed using a draft ISO HDLC procedure was subsequently rejected when put By early 1977, the ISO had still procedure to letter ballot at the link level. The draft ISO and a modified proposal developed. not reached agreement on the portion of HDLC dealing with symmetrical and balanced classes of procedures which v;ere of most interest to X.25 In the unbalanced class of procedure, one station (designated the primary) has total responsibility for control of the link. This class had been strongly supported by IBM whose Systems Network Architecture was, at that time, designed to rely on a large central mainframe controlling dependent Over the opposition terminals. response mode (ARM) the primary. potential The LAP in v>^hich in X. 25(1 of IBM, this class had been extended to permit an asynchronous a secondary may transmit without receiving explicit permission from 976) had been based on the deadlock problems were detected v;ith ARM proposal. However, during balloting the proposaP^, and a new balanced class of 20 i_ procetlures -vas developed. For nu this class, made between dii^tinction is primary and secondary i A version stations, thus giving both stations equal control over the link. asynchronous balanced mode (ABM) was proposed as a revised link mode, known as of this access procedure for X.25, or LAPS. To avoid the meeting risk of further CCITT SG of Eventually a experts and ISO VII TC/97 met compromise "mutually objectionable Despite the compromise, the earlier There was a danger protocol. LAP implementing the old had adopted LAP implementing its the in that both to reviev; their differences. v;as on the books as the official X.25 link access DTE vendors would camps with incompatible continue standards. Already network and by early 1977 had proceeded much too a newly adopted CCITT Approval Recommendations" joint was reached. both parties," netv;ork to convert immediately to Draft 1977 April in network developers and protocol, leading to two DATAPAC to LAP standard preferred standard, of CCITT and ISO standards making, a inconsistencies between procedure ballot letter between To make clear LAPB. LAPS was obtaining for was used Plenaries that ratify to TCTS far in the new "Provisional LAPB; fifteen administrations approved and three abstained.^ At the February 1976 At the first meeting of '*^ fast-select service.'^' and call facility select SG VII SG The meeting VII in it had been agreed still Support and ANSI for a hold datagrams over for further study.^ 1977, the participants examined a proposal from Japan for a fast-select facility would allow the inclusion disconnect packets. Thus for small quantities would provide to data of simple two packet exchange, much -- e.g. like of user data in the call set up transaction data - the fast select a datagram service. However, fast requires the netv;ork to incur the overhead of virtual circuit set-up. for inclusion of a X3S37.'*^' '*'* service that provided IFIP datagram service in X.25 came primarily from IFIP Working Group represented the scientific community and their interest maximum flexibility to the user. The ANSI position in economic terms: i.e. that a network was formulated by a subgroup whose members represented ARPA, NCR, Xerox and Western Union. datagram service v;as couched in 6.1,*^ Their support for datagram service would be far more cost 21 effective for shoit transactioti type traffic than v/uuld conducted and publisfied a survey and rallied permanent additional support virtual circuits February 1960. In SG as a Vli be users and vendors on of ISO TC97/SC5. in means to To buttress virtual circuits. countered position ttiey a datagram interface,''^ ttie desirability of Tlie Frencti ttieir witfi a proposal to use provide a low-overfiead data exciiange capability. adopted a revised version X.25 (X.25(1980)) of v;hicfi included datagram service and fast select as "additional," or optional services as opposed to "essential" parts of the standard. Inclusion of a datagram specification has turned out to be an empty gesture. absence any obligation of has done so. the In implement datagram service, not a single public packet netw^ork operator to X.25(1984) calls for the datagram option to be dropped altogether.'*® The fast select option has been implemented only by NTT."* Lessons From X.25 7. Louis Pouzin has described the X.25 standardization process as a "well engineered political _ __ coup , . A 48 II . small, highly motivated group of packet sv;itching carriers developed a new/ standard breaking time and persuaded a very conservative standards organization, the CCITT, The process was proposal. informal processes used. of a In engineered' with respect 'well 1975, there were a standard before implementation seemed that the most important issues number to stake had at of to its compatible with the emerging networks. telecommunication network providers could secure communication market. X.25 was a 'coup' Rapporteur's group) seemed draft proposal supported to by in the A DTE/DCE manufacturers standard their claim for a sense X.25 was do with technical computer terminal packet accept its timing and the mix of formal and sense. motivation for international telecommunication standardization. A persuade computer mainframe and record- imminent public packet networks; adoption make good little to to in that just was share when the in 'political' in the sense compatibility, the usual standard was needed to to produce equipment also required so that the rapidly growing data institutional processes (the be foundering, the informal group suddenly produced a completed a Furthermore, the group succeeded number in of telecommunication administrations gaining the approval of the CCITT. and carriers. 22 7.1 . Evaluating X.25 Before we can draw any conclusions from the case X.25 for other standards efforts, of first, by the extent of use of the first Success can be measured establish vWiether or not X.25 can be judged a "successful" standard. two ways: we must in standard; and second, by a more subjective assessment of "technical quality". Over 30 countries have public packet switching services -- all X.25 compatible - while only a dozen countries offer circuit switched digital services."*^ Most administrations were able to take advantage of the revised X.25 proposals developed during the 1977-1980 implemented the 1976 version An important goal manufacturers to of the of X.25, and it took until network providers was develop X.25 compatible products: 60 computer and terminal manufacturers made by supporting been X.25'*^. 1981 , IBM finally joined in^ by Telenet as conforming certified vendor. In 1983 CCITT study for Telenet to fully convert to X.25 (1980). induce terminal and computer equipment to in this TCTS A survey of 1978 showed 25 already committed to they have been successful. in and by 1983, more than 100 software packages had to X.25(1980), including software for every major computer Another two dozen vendors supply standalone interface boxes Some 57% of the hosts connected to Thus X.25 is question now is X.25 internally. more ambiguous. Only a few The public packet networks struggled and Tymnet began Telenet did not of the uncertainties to facilities. billion profit until 1983.''^, the of its own. On for number same year of in the other hand, the TRANSPAC characters per month. ^' services is it used? The answer to customers between 1976 and 1980. traffic sessions quadrupling v/hich AT&T finally in volume on Telenet three years. began many users over more than 8000 X.25 connections and Still, offering an X.25 European PTTs, by adjusting the leased lines versus packet services, are gradually forcing France's key users, network operators have adopted surrounding X.25 resolved by 1980, increase rapidly with the show a based packet service tariffs of of the largest private Finally, for internal networks. widely available to the data communications user - but However, with many for host-to-X.25 support. Telenet have user-supplied X.25 interfaces'*^. such as the U.S. Federal Government have adopted X.25 as a standard this period: only four networks relative to public data carries traffic of 100 23 X.25 Fioni a teclinical perspective. still both lias defenders and detractors. its ambiguities of X.25(t976) were resolved by X.25(1980). but multiple options or interpretations. As of X.2.5 has presented somevWiat "^^ alike. 7- 1 ). particularly scathing critique by moving target to and revisions proposed expensive for transaction traffic for in 1976 1984 As noted will to fault for the X.25 AT&T for its for development packet switched networks based on the ISO protocol by IBM put it. was the to ratify "The evolution network implementors and terminal designers to significantly differ between videotex service earlier, in to support HDLC common at the packet level. Pouzin terminal interface for both circuit standard^. The introduction further evidence of the is is packet voice. unv/arrented complexity of a a no public network has offered datagram eliminate datagrams from X.25 altogether. Thus, X.25 and cannot be used such as Pouzin continue argued strenuously solution which was resolved by defining a datagram "option," ended service, clear defeat for datagram proponents. Critics tfie of This poses continuing problems for those writing X.25 host software. datagram battle over service, many cases As a consequence, packet network services continue countries (Table The of a one in Most need of a and new LAPD a simpler terminal for protocol. On balance, X.25 (1976) must be seen as a technically but by no means ideal. Revisions scheduled for adoption weak standard, in 1984 should with X.25 (1980) as better, permit X.25 to be used finally by microcomputers over the dial-up network. 7.2. Lessons X.25 represents only one of the much to read too standards setting Lesson V. many standards into a single case. activities to was computer communications; Nevertheless, draw several lessons from An Imperfect Standard X.25 (1976) for is Better Than hurriedly adopted with known implemented the 1976 version were obligated to we believe it may be it is risky to attempt instructive for future this history of X.25. None ambiguities. modify their Networks - and implementations their customers •• to align with the who 1980 24 Table 7- 1 : Comparison of Differences in X.25 Public Networks Country Germany France U.K. Japan Network Datex-P Transpac PSS DDX Packet Level: 25 version • at significant success. From their succeeded perspective, X.25 equipment manufacturers to in its same networks as most important objective: it a great induced take packet networks seriously and to begin designing compatible By 1975, Tymnet, equipment. Yet X.25 has been hailed by these expense. RETD and Telenet had been established without an international TCTS, the EEC, the BPO and the French PTT had committed themselves standard interface. establish networks with or without a CCITT approved standard. If to these networks had been allowed to evolve independently, the chances of reaching agreement later would have been drastically reduced. X.25 suggests that economic and competitive pressures are forcing the rapid implementation of computer communications and standards. If standards are to that these make new applications will a contribution to these technologies they must be developed on a compatible time scale. The typical eight year development cycle The the current competitive environment. the standard's rigour in abberation: rather, if, alternative, as revision. We refined, is Lesson revisions. The direct The in the future, alternative, to delay the benefits of in is far is too slow in some of to sacrifice is consequence from a commercial perspective, an imperfect standard perceived to be too costly 2: ISO and CCITT would argue that the imperfections can expect more hurriedly developed standards subsequent of suggested by X.25, favour of more rapid approval. sucsequent changes and be implemented with or without of is a need for X.25 were not an better than none, we expended in and more new technology effort while standards are a competitive marketplace. Build a coalition in private meetings with other vendors and submit a joint proposal to formal institutional standards meetings The Rapporteur's group on packet switching involving carriers full reach agreement whereas an informal group TCTS, Transpac, Telenet, BPO and NTT succeeded. This smaller group, containing those committed to the implementation motivation for agreement, met the failed to more than of packet switching, and thus with the greatest economic ten times in 1975, sometimes for as much as a week, while Rapporteur's group was meeting only three times. The informal group then presented the Study Group with a complete draft of a standard at the February 1976 meeting. full 26 Our economic theory of standards suggested a positive feedback model unit eventually forces the standard. economic momentum. Tiie coalition of Xerox. DEC and AT&T network standard, and the moves by American Presentation Level Protocol Syntax Lesson has there Standards agreements 3: been some X.25 (1976) to were of creating sufficient adoption If The leading new technologies with are CBS for its North be most easily reached after likely to and technology, the role in the •- to benefit those 15, it in when a number of had ARPANET to many as 15 under for the 1976 CCITT operation, but as to be developed development was taken by TCTS, Transpac and the BPO. The three RETD, Telenet, and Tymnet spearhead the development -- while agreeing to eventually conform, effort. Conversely, without the experience of the three networks already packet networks the Ethernet local area videotex are examples of similar coalition formation. there were only 3 networks a standard was operation in less motivated to publicized of support by Canada and to solicit for for the one means Organizations whose implementations are imminent are the ones was approved when networks already push is which the dominant develop standards. discussion for 1980. Plenary. Intel to experience practical implementations are imminent. most motivated lor Coalition formation in experience, it would not have been possible develop a standard. The rooted firmly to in operation, plus the widely generate sufficient interest circuit switching tradition in in the telecommunications industry had caused some reluctance during the early 1970s on the part of the PTTs and to the carriers to establish public packet networks. But by 1975 technology to recognize its potential. The fact that four year intervals forced the proponents of a standard to most had had sufficient exposure CCITT Plenary sessions are held only move quickly in at 1975 or face an additional four years delay. The best time some the to develop a standard thus appears to be a during a very narrow window after there operating experience, and field, The need but before these for some when there has been a commitment by other organizations same organizations have committed themselves operating experience before the standard is to is to enter divergent approaches. developed means that the leading 27 may have innovators to British Prestel videotex 4: RETD and Tymnet have been been forced on Ungermann-Bass. a pioneer (Similar redesign has Lesson redesign their systems as Telenet. system •- both pioneers Managers who can commit who got ahead in of the their organizations will obliged to do. and the local area networks, standards making process.) be more successful in securing adoption of a standard than technical experts without decision-making responsibility CCITT processes have from the more political must check with led been characterized by a separation The usual CCITT Plenary meetings. participant of technical only an intermediary is agreeing to any proposal. However, the effort his chiefs before study groups to who develop X.25 was by top management from the key organizations: Norton was Assistant Vice-President (Planning) of the the traditionally Computer Communications Group BPO, Roberts was President Precisely because of the managers took a of TCTS, Kelley was Head of Telenet, and Picard became Directeur General economic importance direct interest. Data Systems Division of the They had the of of in Transpac. standards to these fledgling netv;orks, these authority to make commitments on behalf of their organizations which greatly accelerated the process. Lesson interfaces 5; Standards agreements are more where there is likely to be reached for a natural division in responsibility lor the supply of "natural" interfaces, i.e., equipment or services. Recotnmendation X.25 defines the interface between customer terminal equipment and the communications network. For public data services, customer equipment is generally provided by the user rather than the network provider, thus creating a division of responsibility and a "natural" interface. For "natural" interfaces, both sides can further their Network operators are freed from providing multiple interfaces extent that qsers demand data self interest to by adopting a standard. many equipment providers. To the terminal equipment capable of being used on public netv.'orks, equipment providers gain a larger market if their devices can be used interchangeably in many different countries without modification. By contrast, host-to-host standards division of responsibility and were not -- e.g. finally the ISO's Transport Protocol agreed to until 1983. •• do not define a natural 28 Lesson The layered approach speeds the development of standards. 6: k Recommendafion X.25 separates DTE/DCE the various network functions at the physical layer, a link layer and a packet layer. developed by separate working groups, and theory, layering allows standards for each layer to be In is interface info a expected changes to facilitate standards as they to the evolve. The history of X.25 suggests that layering except that the Study Group adopted a the ISO HDLC proposals. link does help. X.25 (1976) might not have been ready access procedure and a physical and included in the revised version of X.25 published physical level or the packet level. The speed with in 1978. which jointly HDLC this revision link by ISO and the CCITT No changes were was time standard based on the event. ISO subsequently rejected the proposed In A compromise procedure (LAPS) was developed procedure. level in access in 1977 required at either the possible was due largely to the layered structure of the standard. The success of Interconnection, and The cooperation Lesson 7: the in layering concept is reflected in the ISO's model Open Systems for the activities of the IEEE's Local Area Network group, 802. with ISO leads us to our final observation, namely: Standards of higher quality and greater generality are settings organizations cooperate with likely to be produced each other during the development and if standard- drafting of new standards. The HDLC standard was bit-oriented data link control been a strong overlap CCITT SG VII in the result of considerable effort aimed at developing a procedure the for membership to of related committees, the 1977 and the ISO working party charged with developing HDLC joint set a under pressure from the EEC, the ISO and the CCITT were able develop transport protocols in 1982. efficient synchronous communications. Although there has often session between precedent cooperation between the these two standards organizations. Efforts to institutionalize failed until, more to this harmonize for formal precedent their efforts 29 8. Conclusion The need demands on to develop new computer communications standards has placed traditional standards setting processes. reflected this stress, changes to and the process. standard rushed to in Economic agreement. for the first time, priorities The among on the meetings and committed employed of Recommendation X.25 the course of developing a standard, the key participants in through informal meetings The development were allowed to override technical formal, plodding procedures of the the key parties. Top managers their organizations to in such areas as videotex, services can be expected to exhibit many of the same made significant concerns and a weak CCITT were circumvented rather than technical specialists sat an agreement. The principal as well as an unusual degree of cooperation betv/een Current standards efforts unprecedented of layering CCITT and local area networks, or electronic patterns of decision-making. was the ISO. message 30 References 1. Drew, "Analysis Bottaro, of Consensus Standards." Tech. 2. Factors Affecting report No. MIT-EL Braunstein, Yale M.. and Lawrence Economic Analysis", Brandeis 3. 4. Fiorovanle, Janice, "Mirlan Brock, G., J. tfie Demand for and Supply of Voluntary 82-0003WP, MIT Energy Laboratory, 1981. White, "Setting Technical Compatibility Standards: University. Department of An Economics A Standard?," MIS Week. May 1 1 1983, "Competition, Standards and Self-Regulation the in Computer Industry," in Regulating the Product: Quality and Variety. Caves, R.E. and M.J. Roberts, eds., Ballinger, 1975. 5. Wallenstein, Gerd 6. Crane, Rhonda Publishing, 7. Okabe, T., J., D., international Telecommunication The Politics ot International Agreement Oceana, New York, 1977. Standards: France and the Color TV War Ablex Norwood. New Jersey, 1979. "Evolution of Standards Through the ITU (CCITT Recommendations)," P'-oc. ICCC 78. ICCC. 1978, 8. Pouzin, L., "Virtual Call Issues in Network Architectures," European Computing Conference on Communications Network, 1975, pp. 603-618. 9. Pouzin, L., "The Netv;ork Business - Monopolies and Entrepreneurs," Proc. ICCC 76, ICCC, 1976, 10. Davies, "The Services Provided by Datagrams, Virtual Circuits and Host-Host Workshop on Data Communications, International Institute for Applied Systems Austria, September 1975, pp. 43-52. E.W., Protocols," Analysis, 11. Manning, E.G., "On Datagram Service Networks, Vol. 2, No. 2, May in Public Packet Switched Networks," 1978, pp. 79-83. 12. CCITT, White Book ITU, Geneva, 1969, Fourth Plenary Assembiy 13. CCITT, Green Book ITU, Geneva, 1973, 1972 14. Vaughan, Proc. 15. V.J., Jr., ICCC "Standardization 76. ICCC, 1976, in Fifth of CCITT, Mar del Plata, 1968 Plenary Assembly of CCITT, Geneva, December Data Systems with Emphasis on Telecommunications," . Roberts, L.G. and Wessier, B.D., "Computer Network Development to Achieve Resource Sharing," IFIPS Proceedings 16. Computer Scantlebury,R.A. and NCC 1970, IFIPS, 1970, pp. 543-549. Wilkinson, P.T., "The National Physical Laboratory Data Communications Network," Proceedings ICCC 1974. ICCC, August 1974, 17. Radwin, M.S., "Tymnet's Advanced Communications Technology and Recent State-of-the-Art Service Introductions," Proceedings ICCC 1980, ICCC. October 1980, pp. 113-118. 18. Roberts, L.G., "The Evolution of Packet Switching," Proceedings of the IEEE. Vol. 66, No. 11, November 1978, 19. Danet, A. ef a/, pp. 1307-1313. "The French Public Packet Switching Service: Proceedings ICCC 1976, ICCC, August 1976, pp. 251-260. The TRANSPAC Network," 31 20. 'RCP, the Experimental PacKetSwitcheci Data Transmission Service of the French PIT: History. Connections, Control," Proceedmcjs ICCC 1976. ICCC, August 1976, pp. Bacht3. A. ef al. 37-43. 21. 22. "An Overview Proceedings ICC W74. IEEE. June 1974, pp. 2A.1 Norton, D.J. and Bowie. P.G.. Clipsham. W.W., Glave. F.E. and Narraway, of Third International Conference on 23. Dataroute: of System and Performance," 2A.5. • "Datapac Network Overview," Proceedings Computer Communication, August 1976. pp. 131-136. fvl.L.. Hadley. D.E. and Sexton, B.R., "The British Post Office Experimental Packet Switched Service (EPSS) - A Review of Development and Operational Experience." Proceedings ICCC 1978, ICCC, September 1978, pp. 97-102. 24. Barber, D.L.A., "Progress with the European Informatics Networks," Proceedings ICCC 1974. ICCC, August 1974, pp. 215-219. Kato and Y. Yoshida, "A Design of Packet Switching System," Proc. ICCC ICCC, Stockholm. August 1974, pp. 151-162. 25. Hirota, K.. fvl. 26. Takatsuki, T., '78, "Japan's Future Public Network for Computer Communications," Paoc. '74. ICCC Kyoto, September 1978, pp. 177-182. December 27. Kato, M., NTT, Personal Communication, 28. Trans Canada Telephone System, "Datapac Standard Network Access Protocol," Tech. report, 29. 22, 1980 Computer Communications Group, TCTS, November 1974. and James Forgie, "Experience with Speech Communication in Packet Networks," IEEE Journal on Selected Areas in Communications, Vol. SAC-1, No. 6, December Weinstein. Clifford J. 1983, pp. 963-980. 30. Sproule, D.E. and Mellor, "Routing, Flow, and Congestion Control F., Network," IEEE Transactions on Communications. Vol. Com-29, No. 386-398. 4, in the Datapac April 1981, pp. 31. Rapporteur's Group on Packet Switching, "Report of the September 1975," COM-VII-237 32. and L. Kleinrock, "Flow Control: A Comparative Survey," IEEE Transactions on Communications. Vol. COM-28, No. 4, 1980, pp. 553-574. 33. Rybczynski.A., Wessler, Geria, Meeting, Geneva, 16-19 fvl. B., Assessing Data Networks AFIPS, 1976, pp. 477-480. - "A New Communication Protocol for The international Packet-Mode Interface." Proceedings NCC 1976. Despres, Lavandera, 35. iSO/TC97/GCS/WGl. "Statement November 1976)," COM-VII-40 36. GostI, 37. Carlson, D. L., R., and Wedlake, J., Personal Communication, November 27, 1980. 34. 38. Fifth to CCITT (Extract from Document ISO/TC97/SC6 No. 1342 "A Problem with the X.25 Link Access Procedure," Review, Vol. 7, No. 4, October 1977, pp. 37-40. J., (HDLC Rapporteur CCITT, "Circular Letter No. 61 to ," . iSO/TC97/SC6) Interview ACM Computer Communications November 24, 1980 32 39. CCITT. Orange Book: Vlth Plenary Assembly of CCITT. Geneva. September 1976 ITU, Geneva. 1977. COM 40. NTT. "Fast Select Facility, November 1976." 41 NTT, "Fast Select Facility. November 976," COM- VII-30 42. IFIP 43. USA. "Working Paper on Datagram Procedures. November 1976." COM-VII-25 44. USA. "Generic Requirements 45. ANSI. "Results of tlie ANSI X3S37 Survey on Datagram Interface Standard," Communication Review. Vol. 9, No. 1. January 1978. pp. 17-21. 46. Meir, 1 WG6.1, "Possible Extensions Edwin E.. Vol. 12, No. 10. = VII-29 for to Recommendation X.25, December 1976." COM-VII-45 Datagram Service. May 1977," COM-Vil-96 "Packet switcfiing and X.25--Where October 1983, pp. 121-138. to ACM Computer from here?," Data Communications, 47. Matsumoto. M., J. Mizusav/a and H. Ohnisfii. "DDX Packet Switched Technology," Proc. ICCC 78. Kyoto, September 1978, pp. 177-182. 48. Pouzin, Louis, "The Seven Curses of the 80's," Presented at Network and its INFOTECH, London, November 1980 49. Initial Experiences with a McGibbon, C.I., Gibbs, H., and Young, S.C.K., "DATAPAC Commercial Packet Network," Proceedings ICCC 1978. ICCC, September 1978, pp. 103-108. 50. Deaton, G.A. and A.S. Barclay, "IBM finally gives U.S. users Communications = Vol. 10, No. 9, September 1981, pp. 83-93. - ticket to X.25 Networks," Data . 51. Transpac. "Who Offers the World's Largest X25 Network?," Data Communications, Vol. 12, No. 10, October 1983, pp. 237, Advertisement 52. Hess, M.L., M. Brethes and A. Saito, "A Comparison of Four X.25 Public Network Interfaces," IEEE ICC 53. Pouzin, '79 L., Conference Record. IEEE, 1979, "The Case for Revising of X.25," 1976, pp. 143-144. kS33 OH^il pp. 38.6.1-38.6.8. Computer Networks, Vol. 1, No. 2, September 5-^/- MIT LIBRARIES 3 TDflD DDM 461 Tb3 DliQ ^mM^'Z'i ICQf Lib-26-67 ^-^)lp^'e oK "Py^^ G^v/tS^