Chapter 11 - Standards: An Enabling Institution 1979-1984
11.7 ISO/OSI (Open Systems Interconnection): 1981 - 1982
In November 1980, the sigh of collective relief that the OSI Reference Model had advanced to the DP stage reinvigorated those seeking to influence the various protocol designs. For the members of ISO TC 97/SC16 the definition of the transport layer protocols, especially the need for an end-to-end datagram protocol, became a priority. CCITT already had made its demands known when it approved its transport layer standard for teletext - Recommendation S.70 - several months earlier.28 In January 1981, ECMA standardized on the five classes of transport protocols it had earlier recommended to both ISO and CCITT: class 2 being equivalent to CCITT’s S.70 and class 4 being equivalent to TCP.29 Predictably, these actions influenced the proceedings of SC16 that prized agreement with both CCITT and ECMA and joint meetings ensued.
With the Reference Model advancing to the DP stage, it’s near certain standardization induced a flurry of commercial activity. At the National Computer Conference (NCC) in May, data communications took center stage for the first time; with seminars such as “Transport and Session Protocols in the Context of the ISO Reference Model” and “Local Networks and the Ethernet in Particular.” Luminaries such as Cerf, Greg Hopkins, David Potter, and John Day gave talks.30 Articles on OSI (“Coming of Age: A Long-awaited Standard for Heterogeneous Nets,”31 and “Reality and the Proposed OSI Standard”32) began appearing in technical and industry publications. And the undeniable importance of the connection protocol X.25 gained support with first AT&T in January and then IBM in June announcing products. No wonder Zimmerman had felt the pressure to complete work on the Reference Model so protocol standards work could begin: market forces were already at work.
In January 1982, less than a year after circulating the Reference Model DP 7498 to its members for passage, TC 97 approved the Reference Model as a Draft International Standard (DIS) and forwarded it to ISO for vote by its full membership. With virtually all of the uncertainty about the Reference Model becoming a standard eliminated, (notwithstanding the continuing criticism from the U.S. delegation that the Reference Model should be a technical document and not a standard) both committees, SC6 and SC16, felt pressured to begin producing the protocols needed to actually build products. To do so meant creating two standards – Services and Protocols – for each layer.33
In June 1982, SC16 at its Tokyo meeting acted first when it approved as Draft Proposals (DPs) both transport layer Service (ISO/DP 8072) and Protocol (ISO/DP 8073) standards. The U.S. (ANSI) once again voted “no” – citing incompatibility among the five classes of DP 8073 as well as the claim that the DP could not be implemented.34 Efforts to resolve the U.S. objections, as well as objections of others, prior to forwarding the DPs to TC 97 took on a new level of seriousness. Modifying TP-4, the virtual circuit protocol over connection-less networks, was assigned to the NBS; virtually guaranteeing a protocol similar to TCP. Despite the differences, the quick action on DP 8072/73 reflected real cooperation with both CCITT and ECMA. (After all, both CCITT and ECMA got what they wanted and more: class TP-0 providing minimal functionality, consistent with a connection network layer such as X.25, and class TP-4, providing functionality comparable to TCP, sought by the computer community.)
On the other hand, progress within SC6 to create a connection-less protocol at the network layer stalled while the committee completed its connection-based Service and Protocol standards. In part, progress lagged because the Reference Model did not yet call for connection-less communications. What motivation would a connection-oriented committee have to create a connection-less protocol?
By the fall of 1982, Reference Model DIS 7498 votes began coming in and the U.S. once again voted no.35 While the three objections that accompanied the no vote were the same ones the U.S. had attached to prior yes votes, the fact that the U.S. – the secretariat –voted no seemed divisive.36 In a December 1982 Data Communications article titled: “Network users group shocked as ANSI votes against OSI model,” Sheldon Blauman, founder of the Network Users Association (NUA) was quoted:37
It is both a surprise and a shock to me. A vote against Open Systems Interconnection is like a vote against apple pie and motherhood. I see the situation as rather frightening.
Even though the final vote on DIS 7498 tallied 22 yes, 2 no, and 1 abstention, declaring 7498 an International Standard remained on hold while an effort was made to resolve the US objections. (Voting no with the U.S. was Germany.)
At year-end 1982, the promise of OSI to bring order to the conflicting technologies within computer communications remained just that, a promise. The overarching OSI architecture of Reference Model DIS 7498 seemed close to becoming an international standard. But since DIS 7498 assumed a connection paradigm, accommodating the connection-less world of LANs meant a Reference Model addendum – a document modifying an existing standard – although being worked on, had to be voted through the same DP, DIS, ISO process.
In one encouraging note, SC16 passed a TP-4 version of DP 8073: a virtual circuit protocol over a connection-based network. Yet until SC6 passed a connection-less network layer, LAN products using ISO OSI standards remained only a distant possibility. What were LAN manufacturers to do in the meantime?
At year-end 1982, two connection-less, or datagram, communication protocols for LANs did exist: TCP/IP and XNS. Their success would impact the eventual outcome of OSI.
- [28]:
“Teletext: an office-oriented text-transmission service,” Data Communication, Jan. 1981, p.68.
- [29]:
Standard ECMA 72, January 1981
- [30]:
J. Peter Schmader, “Computer Conference Sharpens Focus on Data Communications,” Data Communications, April, 1981, pp. 129-132
- [31]:
Harold C. Folts, “Coming of Age: A Long-awaited Standard for Heterogeneous Nets,” Data Communications, Jan. 1981, pp. 63-73
- [32]:
Bert D. Moldow, “Reality and the Proposed OSI Standard,” Data Communications, June 1981, pp. 77-80
- [33]:
Peter F. Linington, “Fundamentals of the Layer Service Definitions and Protocol specifications,” IEEE, Dec. 1983, pps. 1341-1345. The Service standard sets forth the capabilities delivered to the user of the layer, the next higher layer, and the constraints on the layer immediately below. (Hence, the Service standards for the transport layer specify the constraints on the network layer and include capabilities for users of both the transport and session layers.) Protocol standards specify how products worked to delivery the Service capabilities.
- [34]:
A. Lyman Chapin, “Computer Communication Standards,” Computer Communication Review, Jan. 1983. p.40-43.
- [35]:
This is after the U.S. had abstained at the DP stage vote.
- [36]:
A. Lyman Chapin, “Computer Communication Standards,” January, 1983
- [37]:
Data Communication, Dec. 1982, p. 50. NUA members included Ford, Bechtel, Grumman, TRW and McDonnell Douglas.