Chapter 08 - Networking: Diffusion 1972-1979
8.9 Metcalfe Joins the Systems Development Division of Xerox 1975-1978
Despite his success, Metcalfe was frustrated within a strictly research environment. He wanted the energy and excitement of engaging in leading edge research, but he also wanted the satisfaction from succeeding in putting ideas to use. He wanted to be more than an engineer. In November 1975, he left Xerox PARC for Transaction Technology, Inc. an advanced product development organization of Citibank. In just seven months, he was wooed back to Xerox by David Liddle to join the newly formed Systems Development Division (SDD).37 SDD’s charter specifically called for the commercialization of the Altos technologies, including Ethernet.
Commercializing the Altos technologies required more than simply selling what existed. The entire system, including Ethernet, had to be reengineered to improve performance, reduce costs and made ready for manufacturing. The initial goal for Ethernet was to increase performance from 3 megabits per second to 20 megabits per second and commensurably improve PUP.38
Liddle remembers the controversy surrounding the decision to increase the throughput, or bandwidth, of Ethernet to 20 megabits per second:
The people at PARC complained about this a lot, and said: ‘Why isn’t three megabits good enough?’ We simply said: ‘We didn’t think three megabits would be good enough over this time horizon,’ because we thought there would be more movement of big files and databases and printing big images, and the product was really virtually playing almost the role of a bus, not an old-fashion ‘beep, beep, beep’ communication line. This was somewhat controversial because it increased the cost and subtlety of designing some of the components.
By late summer 1976, Metcalfe needed to find someone knowledgeable with the latest in protocol developments, including TCP, to lead the redo of Pup protocol. Metcalfe understood the design of Ethernet inside and out, but felt like an outsider looking in when it came to protocol development. Fortunately just such a experienced person existed, Yogen Dalal. Metcalfe knew Dalal and held him in high regard. Dalal would be completing his Ph.D. under Cerf at Stanford in early 1977.
Closing Dalal would not be easy for he also was being strenuously recruited by Kahn at DARPA, Tomlinson at BBN, Taylor at PARC and Steve Crocker at Information Sciences Institute. What gave Metcalfe an edge was Dalal’s abiding interest in communication protocols. In 1975, he had been a member of Cerf’s team coding and testing TCP. Dalal was also a member of the TCP Working Group and deeply involved in the effort to architect a second version of TCP to reflect what had been learned in the testing and use of the first version.
Independently, in September 1976, Dalal’s mentor, Cerf, suddenly resigned both his teaching position at Stanford and Chairmanship of IFIP Working Group - 6.1 to join DARPA where he would manage the packet communication technologies, the Internet project and the network security program.39 Cerf’s impact on networking would continue both through his own contributions and those of his many students, including Dalal, Carl Sunshine, Richard Karp, Jim Mathis, Ron Crane, Darryl Rubin, John Shoch and Judith Estrin, the daughter of Jerry Estrin, Cerf’s thesis advisor at UCLA.
With Cerf at DARPA, Dalal decided to strike out on his own independent of Cerf. In the end, Metcalfe’s offer to create the successor protocol to Pup was simply too good to resist. After joining SDD in 1977, Dalal began assembling his team – including Will Crowther, once a critical member of BBN’s Arpanet team, and Hal Murray – and conducting an in-depth review of Pup.
Metcalfe also recruited James White, whom he had known from his Arpanet days and who was now working at Stanford Research Institute, to become Manager of the Communications Software Group in 1977.40 In addition to working on the rearchitecting of Pup, White’s Group assumed responsibility for commercializing a PARC developed, distributed-Ethernet-based electronic mail system called “Grapevine.”41
By early 1978, with Ethernet working and product sales no closer than when he had joined SDD, Metcalfe found himself frustrated and restless. Wanting to see his Ethernet technology commercialized before others exploited the opportunity, he argued that Xerox should sell Ethernet products unbundled from computer systems. However, management did not see it his way. In the spring of 1978, Metcalfe issued management an ultimatum, with the veiled threat that he would resign unless Ethernet be made available for sale. He remembers:
When I came back to Xerox in 1976, we were about two and a half years from product shipment and in 1978 we were about two and a half years from product shipment. And my analysis, at the time amateurish, was that there were things other than engineering that you needed to do right to succeed, and, apparently, since we were so good at engineering, the problems must be in marketing or manufacturing or something, and so I wanted to find out more about those, because I hated failing and we were – WE, not they, WE were failing. Since I was buried in an engineering organization, I gave Xerox seven months notice and said I want to at least report to a general manager.
Metcalfe did not get his wish, and true to his word, left Xerox at the end of 1978 to become an independent consultant.
- [37]:
SDD was created in January 1975.
- [38]:
Technically, the higher speed LAN was called the X-Wire - X for Xerox. It would, however, become the higher performance Ethernet and the X-Wire naming would be abandoned.
- [39]:
By late 1976, the ARPA Packet Radio Network, ARPANET and the Atlantic Packet Satellite Network were connected together using two gateways between the Packet Radio Network and ARPANET and three gateways between the ARPANET and the Packet Satellite Network. The gateways, the store-and-forward packet switches that enable the networks to be interconnected, signaled the emergence of a new class of devices performing internetworking.
- [40]:
White had been recruited to SRI in 1972 by John Melvin, a system programmer whom he had met at NWG meetings. Jon Postel, who joined SRI after receiving his PhD. from UCLA and, after a brief detour through MITRE in Washington D.C., worked with White. (During this entire period, Postel retained responsibility for the Request For Comment’s (RFC’s); facilitated by his work at SRI, the Network Information Center (NIC), and repository of all on-line RFC’s.) White and Postel worked on a family of protocols that would allow computer procedures to span multiple networks.
- [41]:
White would become a key contributor to the standardization of email with ISO.