Chapter 09 - Networking: Emergence 1979-1981
9.5 Robert Metcalfe and Digital Equipment Corporation
While consulting to LCS, Metcalfe contacted Sam Fuller and Bill Johnson at DEC and was retained to work on their ten-year long-range plan (especially in the area of connectivity). Immediately he met with Gordon Bell, Vice President of Engineering at DEC.
Bell, like Poduska at Prime, had experience with Arpanet, including attending some Network Working Group meetings. He took it as a given that computer connectivity was important to DEC. Only he differed with those who had architected Arpanet. Bell explains:
The place I disagreed with ARPA was whether hosts could or couldn’t be IMPs. All the DEC work was based on the notion that a host could be an IMP. We were building IMP-sized computers, so I didn’t care whether you actually computed or did switching and computing. Everybody said: ‘No, you can’t ever have switching and computing in the same box.’ ‘We’re not going to force our users to have a whole switching network of IMPs around. This is crazy.’
Bell’s first efforts to launch DEC into communications began modestly. In early 1974, he decided to try using a PDP-11 as a new remote job entry station for their DEC System 10. He assembled a three-person team, one engineer each from the PDP-11 and DEC 10 engineering groups and Stu Wecker, his communication expert. Wecker had joined DEC in June 1972 as a researcher and had interconnected three minicomputers. Known as SHARP, for Stu’s Homogenous Asynchronous Re locatable Process operating system, the demo so impressed Bell that when he needed a communications expert, he turned to Wecker; despite the fact that before joining DEC, Wecker had absolutely zero experience in computer communications: “I didn’t even know the word ‘protocol.’” In seven weeks the team designed a new communications protocol: DDCMP, Digital Data Communications Message Protocol. DDCMP separated control of the devices at the end of the communications line from control of the communications line itself; a division between the “host-to-host” and network responsibilities and thus a two-layered protocol.
By the beginning of 1975, customer demand replaced management vision as the driving force motivating DEC to interconnect their computers. At first the need came from industrial and laboratory customers, requests that Bell forwarded to Wecker, their “guru on networking.” Wecker began talking to customers and soon headed up a small networking committee to determine the specifications for DEC’s computer communications network, to become known as DECNET.
Knowing little about networking, Wecker studied material on the Arpanet. Of all the papers and articles he read, two proved essential to DECNET: Bob Metcalfe’s PHD Thesis and “Communication Networks for Computers” by Donald Davies and Derek Barber. The Davies and Barber book provided the concepts of layering, symmetry, and peer-to-peer networking (a key differentiation from IBM’s SNA). The writings of Louis Pouzin, Hubert Zimmerman, Vint Cerf, and Carl Sunshine also influenced his thinking.
DECNET shipped in late 1976, but, due to lack of functionality, such as terminal protocols, customers reacted poorly. Members of top management worried that maybe networks were just too complicated, and that DEC should concentrate on minicomputers. But Bell, committed to the importance of the VAX computer project launched in 1975, would have none of it, convinced that communications were essential to computers and that to achieve the VAX goal of a single architecture spanning the entire spectrum of performance would require multi-processor communications.
Fortunately, Bell had maintained his contacts with academia. When Metcalfe and Boggs submitted their paper to the ACM in May 1975, Bell was selected as one of the reviewers. Sensing its implications for DEC, he circulated the paper within DEC, nearly a year before its publication in July 1976, and initiated a number of low level networking developments. He no longer questioned whether networking would play a role in DEC’s future, he just wasn’t certain how or when. Customer demand, meanwhile, kept growing for computer communications, and DEC delivered DECNET Phase II in 1978.
Bell completed documenting his vision of DEC’s product strategy in 1978. Known as the Distributed Computing Environment (DCE), it was approved by the DEC Board of Directors in December 1978. (Subsequent evolution is shown in [ ] brackets.)3
Provide a set of homogeneous, distributed-computer-system products based on VAX-11 so a user can interface, store information and compute without re-programming or extra work from the following computer’s system sizes and styles:
- via [a cluster of] large, central (mainframe) computers or network;
- at a local, shared departmental/group/team (mini) computer, [and evolving to a minicomputer with shared network servers];
- as a single user personal (micro) computer within a terminal [and evolving to PC Clusters];
- with interfacing to other manufacturer and industry standard information processing systems; and all interconnected via the local area Network Interconnect, (NI)
[i.e. Ethernet] in a single area, and the ability of interconnecting theLocal Area Networks (LANs) to Campus Area and Wide Area Networks.
Consequently, Bell warmly welcomed Metcalfe’s involvement with DEC in 1979; especially as none of the experimental projects he had launched after reading Metcalfe’s ACM paper seemed likely to result in the Network Interconnect needed for the DCE. Bell recalls:
I didn’t know what the hell I was going to do. We didn’t have any good physical things. I didn’t want to go off and make them. We were in the throes of deciding between…There was the ring crowd. The ring looked good to me, but…
Bell was familiar with token ring and knew that Prime and a start-up of ex-Prime employees, Apollo, were both committed to using token ring. Furthermore, Saltzer and Clark had told DEC that they intended to interconnect the gifted VAXs with their version of token ring. Even so, Bell knew of no large operational token ring networks, whereas Xerox operated a large Ethernet network. But how could he loosen Xerox’s ironclad control over their technology? Bell again:
We had to have it, and we would have invented our own. We had two or three different schemes, and I was just turning to that problem when Bob walks in the door and says: ‘Would you be interested in a collaborative effort with Xerox?’ A meeting was held with Metcalfe in a corner room in Parker Street. Parker Street was a prison-like building, a big concrete building. They had been meeting all day when I came into the meeting and said: ‘OK, where are you guys on this thing?’ They said something like: ‘Well, we’d really like to do something, but it isn’t clear..’ I said: ‘ How do we get this thing going?’ They said: ‘Well, you’d better write a letter to so and so.’ I said: ‘This thing isn’t moving fast enough. Tell me what the letter should say, and furthermore, we will write the letter right now.’ So I went next door to a word processing system, and Metcalfe and I composed a letter right then. I wrote: We want to do some kind of a joint venture to get a local area network that would be public, a network that would be used by both companies. I viewed Metcalfe as the behind-the-scenes oiler, greaser that made all of this happen.
The letter was sent to Liddle, George Pagent and James Campbell.
Meanwhile, Metcalfe had begun consulting to IBM. He learned that they leaned towards token ring but had not made any final decisions as to the importance of local area networking or whether to introduce product.
Gordon Bell, Toward a History of (Personal) Workstations (Draft)