Chapter 08 - Networking: Diffusion 1972-1979
8.1 Commercializing Arpanet 1972 - 1975
In 1971, Larry Roberts recognized a working Arpanet posed a new problem: ARPA lacked legal authority to operate a communications network. A new custodian, presumably a private organization, would have to be found to take on the responsibility without sacrificing the military’s preferential use and control of the network for its on-going research. Roberts approached AT&T and BBN. AT&T rebuffed his inquiry and BBN displayed little interest.1 Believing he had time to solve this problem, Roberts buried himself and his small staff in the many other IPTO projects such as artificial intelligence, speech understanding, various computer projects including Illiac IV, and other communication projects like ALOHAnet. Why? When viewed through a budgetary lens, Arpanet barely made the list of ARPA’s important projects. After four years an estimated $3.5 million had been spent buying IMPs from BBN to construct the Subnet.2 (Total investment including the costs of attaching Host computers to the Subnet and IPTO management is an estimated $10 million.) 3 During those same four years, ARPA’s total spending was $833 million; making Arpanet one half of one percent (.5%) to one and one half percent (1.5%) of ARPA’s budget; in other words barely significant.4
Until Arpanet could be divested, however, it required continual expansion and improvement. By June 1974, the number of Arpanet nodes totaled 46.5 The following table summarizes the growth in nodes by user type. (See Exhibit 6.0 Arpanet Nodes by Type By Date.) Government agencies - primarily the military – made up most of the growth between 1972 and 1974. With military usage came all the rhetoric of national security that further complicated divesting Arpanet to the private domain.6
Roberts had been around technology development long enough to know that operating and improving a system that in the end would have to be divested was a headache avoid. So in late 1971 he pressed BBN to step up to the plate to solve his and ARPA’s dilemma. However, BBN saw itself as a government contractor, possibly the seller of product into commercial markets, but never as an operator of a communications network. The mere thought of dealing with the FCC, beginning with the effort to attain approval to operate a network without the burden of regulation, was a daunting barrier to entry. Yet in order to be seen as responsive to Roberts, BBN management created a committee to study the issue.
Exhibit 8.1.0 Arpanet Nodes by Type By Date
|As of Date||Number of Nodes||Universities||Government Contractors||Government Agencies||Companies|
The institutional noose forcing Roberts to divest Arpanet tightened considerably on March 23, 1972, when a Department of Defense (DOD) directive changed the agency’s name to the Defense Advanced Research Projects Agency (DARPA) and made it a civilian-directed agency reporting directly to the Office of the Secretary of Defense.7 It also limited DARPA’s responsibilities to the conduct of ”basic and applied research and development for advanced projects.” And although never a practice, DARPA was precluded from creating separate Research & Development (R&D) facilities. In addition, limits were set as to the maximum funds any one R&D organization could be awarded in a year. Fortunately, Roberts had already anticipated the hidden-yet-clear message to divest Arpanet; although he lacked a solution other than hoping BBN would come to the rescue.
Roberts’ plans to transfer Arpanet suffered a setback when three members of the BBN committee BBN studying whether BBN wanted to commercialize the Arpanet became so frustrated they left BBN and announced they were going to start a company offering data communication network services. Their product would be “conceptually similar”8 to Arpanet. They saw the Arpanet technologies as public property, having been funded with public money, and presumed they would be given rights to the technology because it “theoretically should be available to us and to others.”9 In fact, they even told people that they had “been reassured informally of this by the Defense Dept.”10
BBN’s management felt pressured to act. Needing someone to lead the effort, they began recruiting Roberts to leave IPTO and become President of the Telenet Communications Corp. (Telenet) subsidiary they had created in December 1971. BBN’s timing could not have been better. Roberts had decided already that once the fate of Arpanet had been resolved, it would be time for him to leave IPTO, having already stayed longer than he would have ever expected and longer than was the cultural norm. So Roberts, in his inimitable style, dove into analyzing the opportunity and concluded $10-20 million dollars of computer-time usage resulted in an “optimally cost-effective” nationwide network of roughly fourteen moderately sized computers.11 Convinced of Telenet’s economic feasibility, Roberts accepted BBN’s offer. But he needed time to avoid any appearances of “conflict of interest;” especially with regard to divesting Arpanet. For help, he suggested that RAND, the research organization, study the alternatives and make a recommendation.
On being offered the assignment, Keith Uncapher, the Director of the Computer Sciences Division of RAND, immediately thought of involving Paul Baran. Baran had left RAND in 1968 to found the Institute for the Future to study the question of how society solved problems with long lead times.12 Baran readily agreed to consult for RAND. RAND management then realized the March DOD directive had set limits restricting how many dollars could be given to a research institution and RAND had reached their limit. Baran remembers:
So I said; ‘Oh, what the heck. I’ll start a company, if you don’t mind, RAND, and do the study, if DARPA doesn’t object, and DARPA wants it.’ So that’s where Cabledata Associates first got started, doing a study on the divestiture of the Arpanet.
Baran contacted Cerf, who had just started teaching at Stanford University only a mile down the road. Cerf also took little convincing to help Cabledata Associates (CDA). With Baran serving as principal investigator and Cerf as project scientist, work began by the summer of 1972 with a report due by January 1974.
Yet no sooner had CDA started work, than the three ex-employees of BBN, Lee Talbert (President), Ralph Alter (Operations vice-president) and Stephen Russell (Engineering vice-president), founded Packet Communications, Inc. (PCI) in July 1972. They announced they had venture-capital backing, had secured rights to the Arpanet technology, and would be submitting their application to offer data communication services to the FCC.
By mid-September, Roberts, frustrated from not having found his replacement and feeling the pressure to join Telenet, took J.C.R. Licklider up on his casual offer to take Roberts’ job if he could not find a successor. On October 1, 1972, Roberts announced his resignation from IPTO to become President of Telenet and, on the following day, Telenet filed its application with the FCC. Three months later, PCI submitted their application. Both applications would be approved as regulated common carrier undertakings even though they were going to lease lines from AT&T, itself already a regulated common carrier.
A new market was forming. At first the focus would be selling data communication services. In the future, firms would sell packet switching products.13 Packet switching services and products would prove less important than another market that would emerge from the new paradigm of packet switching - local area networking - the larger story being told in this chapter. As for PCI, it never raised sufficient venture capital and faded away, Telenet became a public company in December 1977 with a market capitalization of $20.6 million.14 On June 13, 1979, General Telephone Company, the second largest telephone company, bought Telenet for approximately $58 million.15 BBN received $13.9 million for their cumulative investment of $5.3 million.16/a>
By the time CDA submitted its recommendation on January 14, 1974, the horse had clearly left the barn. Even so, and reflecting CDA’s deep concern if a few firms came to dominate packet switching as an oligopoly, they recommended a new institutional form, a consortium, that would provide “three essential functions:
- A Clearinghouse mechanism for transferring payments among cooperating entities.
- A mechanism for creating and enforcing common industry standards.
- A mechanism to allow continuously free and open entry, to avoid the formation of any closed oligopolistic structure that will demand close governmental supervision or regulation.17
The DOD had its concerns as well and after another year of meetings exploring alternatives, including the recommendations of CDA, the management of Arpanet was transferred from DARPA to Defense Communications Agency (DCA) on July 1, 1975.
Lawrence G. Roberts, “The Evolution of Packet Switching,” Proc. of the IEEE, Vol. 66, No. 11, Nov. 1978, p.1310
ARPANET Management Study, Prepared by Cabledata Associates, Inc., Jan. 14, 1974
A figure estimated by some of those interviewed.
Burton I. Edelson and Robert L. Stern, The Operations of DARPA and its Utility as a Model for a Civilian ARPA, The John Hopkins Foreign Policy Institute, Nov. 1989
Completion Report, Prepared by F. Heart, A. McKenzie, J. McQuillan and D. Walden of Bolt Beranek and Newman, Inc., January 4, 1978, p. III-91
An example of an on-going network improvement was the redesign of the flow control mechanism in 1972. Baran report, p. 9
Burton I. Edelson and Robert L. Stern, “The Operations of DARPA and its Utility as a Model for a Civilian ARPA,” The John Hopkins Foreign Policy Institute, Nov. 1989, p.6
Datamation, March 1973, p. 122
Larry Roberts, “Network Rationale: A 5-Year Reevaluation,” Proceedings of COMPCON, 1973, pp. 3-5 His analysis also showed that communication costs as a percentage of computer costs would stabilize at eight percent.
Baran received backing from the Ford Foundation and, later, the Arthur D. Vining Foundation.
The interested reader is referred to: Lawrence G. Roberts, “The Evolution of Packet Switching,” Proc. of the IEEE, Vol. 66, No. 11, Nov. 1978, p.1307-1312 and Marvin A. Sirbu and Laurence E. Zwimpfer, “Standards Setting for Computer Communications: The Case of X.25,” IEEE Communications Magazine, Vol. 23, No. 3, March 1985, PP. 35-45.
BBN Annual Report 1979, p. 22
Roberts: “Well, GTE got interested in getting into the business, and so they came after us and they told us that they had learned their lesson with several other companies they had bought and they would never do it again. They’d let us do our own thing and everything would be fine. That lasted for one year and then after that the management changed every few months and they kept on trying to figure out how to position it within their organization”
ARPANET Management Study, Prepared by Cabledata Associates, Inc., Jan. 14, 1974, pp. iv