Chapter 11 - Standards: An Enabling Institution 1979-1984
11.5 ISO/OSI (Open Systems Interconnection): 1979 - 1980
When last observed at year-end 1979, the International Standards Organization Technical Committee (ISO/TC 97) had just approved the OSI Reference Model as a Working Draft (WD) and had authorized Subcommittee 16 (SC16) to make the requested changes and resubmit it as a Draft Proposal (DP). Embedded in the Reference Model WD was the fundamental assumption of communications by making connections, not launching datagrams; this notwithstanding the fact that the leading LAN technologies used datagrams, not connections, to communicate between devices.17 Since the Reference Model WD failed to embrace datagram protocols, it could not satisfy the rapidly growing LAN community. Excerpting language from the introduction of the eventual standard:
The assumption that a connection is a fundamental prerequisite for communication in the OSI environment permeates the Reference Model and is one of the most useful and important unifying concepts of the architecture which it describes.18
In OSI terminology, physical and virtual circuits were addressed by connection protocols whereas datagrams were considered connection-less protocols. Physical or virtual circuit architectures create connections before transmitting data whereas datagram architectures simply transmit data without establishing connections, hence, connection-less. The major LAN technologies, such as CSMA/CD and token passing, were connection-less and therefore were unaddressed by the Reference Model WD.
The connection-based Reference Model satisfied the PTT’s, but for those wanting reliable end-to-end communications over the public data networks or LANs, it meant creating:
- a transport layer protocol providing reliable end-to-end service19
- a change to the Reference Model to allow connection-less communications
- a connection-less network layer protocol
The omission of connection-less communications from the Reference Model WD concerned many European computer scientists as well as Rosenthal and others at the National Bureau of Standards (NBS) committed to having the U.S. Government adopt the eventual OSI standards. But since NBS’s pioneering network, NBSnet, used a CSMA/CD, or connection-less, protocol. If OSI did not include connection-less protocols resembling those evolving in the U.S., NBS would have to abandon its goal of adopting OSI standards. In response, NBS pursued a multi-prong strategy. First, NBS sought help from the American National Standards Institute (ANSI), the U.S. representative to ISO. Only, ANSI’s members, companies, favored proprietary technologies and questioned the economic value of international standards.
NBS next approached the European Computer Manufacturers Association (ECMA). ECMA, although not a formal member of ISO, attended SC16 meetings by invitation. ECMA shared NBS’s concern regarding the lack of OSI connection-less protocols for LANs. For years, ECMA had monitored or participated in INWG and IFIP discussions regarding the need for datagram protocols.20 So ECMA welcomed the NBS as an ally and invited NBS representatives to ECMA meetings.21
John Heafner, head of NBS’s Systems and Network Architecture Division starting in January 1979 and responsible for all NBS networking activities, including OSI protocols, remembers wanting OSI to adopt transport protocols similar to the DOD TCP protocol:22
So through ECMA and other sources, we pushed the definitions written in ISO-ese of the DOD protocols, in particular Class IV Transport, which is the match with TCP. Our thrust from the beginning was to work in ISO, and to some extent CCITT, to make sure that government needs were covered when those protocols were being developed.
In early 1980, knowing time was of the essence, ECMA submitted a draft proposal to both ISO and CCITT recommending four classes of transport protocols, ranging from a minimal protocol for connection networks to a TCP-like protocol for connection-less networks. (All classes would function over a connection-oriented network layer.) ECMA, however, needed an inside champion to argue their cause.
In the spring of 1980, ECMA’s champion emerged. The French PTT, influenced by the changing regulatory environment in the U.S. - i.e. deregulation and the Computer Inquiries - and their growing awareness of the merging of computers and telecommunications, recruited Zimmerman from IRIA, a data processing research institute, to join Centre National d’Etudes des Telecommunications (CNET), the research institute for the French PTT - its Bell Labs. Zimmerman’s new role ideally positioned him to bridge the differences separating ISO and CCITT. He reflects:
One of the reasons for moving was the will of the French PTT to put together, or to get more, of the data processing culture. It was agreed that I would keep participating in standards, as I had done before, and from that time was in a much better position to act as a go-between ISO and CCITT. So I was, within ISO, still in charge of the OSI reference model group, and people knew that I had moved to the PTT’s. They could see that it did not change my way of managing and pushing things, and it was clear that it was supported by the French PTT.
In August 1980, facilitated by Zimmerman as well as the more receptive attitude within CCITT to data processing-oriented standards, CCITT and ISO jointly announced tentative support for ECMA’s transport protocol proposal.
Zimmerman’s next challenge was managing the approval of the Reference Model at the upcoming SC16 meeting in Berlin. Success could not be assumed. Zimmerman knew the current version of the Reference Model had problems. How could it not? Designed as it was by an international committee in record time. Some problems were merely cosmetic, such as the correct use of the English language. Other problems were more serious. The U.S. delegation, for example, repeatedly questioned the fundamental objective of a Reference Model. They lobbied that the Reference Model be circulated as a simple technical report rather than passed as a standard with permanent consequences. Zimmerman thought the Americans undervalued the benefit of making the Reference Model hard to change. (He also recognized the ambivalence among the Americans; many whom would prefer the market decide technological issues, not a public body.)23
The problem that concerned Zimmerman most was whether the right balance between “time to market” and technological elegance had been achieved. Many people, including Zimmerman, recognized that a collaborative standards process did not result in a perfect technical answer. However, for a standard to work in the real world required widespread acceptance despite technical weaknesses and conflicting requirements. Zimmerman recalls:
At that time, somehow, we had to admit, and I was convinced pretty early, that we shouldn’t have something which was far from being perfect from a technical point of view, but had to be accepted and endorsed by the major players. We put a fair amount of pressure for, at some point in time, just freezing things, and saying: ‘Unless there’s a very good reason to change something, we won’t change it.’
At the SC16 meeting in Berlin in November 1980, an air of excitement fueled by nearly three years of hard work energized every crowded meeting room. As the estimated 200 attendees eagerly awaited the vote to approve the Reference Model as a DP, each country delegation formally presented their review of the Reference Model and recommended, or not, its passage. John Aschenbrenner of IBM led the U.S. delegation that again argued that the Reference Model should be approved as a technical report, not as a standard. The dull thud that followed hardly had time to reach the back of the room when Mike Purton of the British Standards Institute gave, as Bachman would recall: “his infamous speech,” condemning the Reference Model for its misuse of the English language.
Zimmerman remembers the backlash to Purton’s speech reporting that the U.K. delegation:
Had given this document to their experts, and they had said unanimously that this document should be put in the basket. That was the end of it, because telling this to a bunch of people who had been working hard as hell to produce something, who know that it’s not really perfect but it has some value, telling them that it’s just good for the paper basket, that was the end of it. Everybody agreed. The U.S. said: ‘We’ll just abstain. We’ll not oppose,’ and that is the way the OSI Reference Model moved to the DP stage.”
Making the OSI Reference Model a DP effectively ordered the layering of computer communication protocols even though OSI had yet to create an actual protocol standard.
- [17]:
Datagrams threatened the paradigms of both circuit-switching and virtual-circuits networks
- [18]:
Information processing systems– Open Systems Interconnection – Basic Reference Model, Addendum 1: Connectionless-mode transmission, Ref. No. ISO 7498: 1984/Add.1: 1987(E)
- [19]:
Hubert Zimmerman, “OSI Reference Model – The ISO Model of Architecture for Open Systems Interconnection,” IEEE Transactions on Communications, Vol. Com-28, No. 4, April 1980, pp. 425-432
- [20]:
INWG (International Networking Group) and IFIP (International Federation of Information Processing)
- [21]:
Bachman:” ECMA was very important, because ECMA was working on their own drafts between our meetings, and ECMA tried to provide a power equivalent to IBM.”
- [22]:
Heafner had been responsible for connecting RAND to the ARPANET.
- [23]:
Data Communcations, Aug 80 pp. 31-32.