Chapter 11 - Standards: An Enabling Institution 1979-1984
11.6 TCP/IP and XNS: 1979-1980
Throughout 1979, Kahn and Cerf lobbied the DOD to make TCP/IP an official communication protocol standard. In January 1980, after a year of code testing, Postel posted the two new protocols, TCP and IP, as Request for Comments (RFCs) and simultaneously as DOD standards.24 Once published as standards, the hard work of converting all networks, including ARPANET, to TCP/IP began. The date of January 1, 1983 was set for the cutover of ARPANET to TCP/IP.
The implications of these developments were not lost on Rosenthal and others at NBS who had decided earlier to standardize on OSI protocols for the Federal Government. Since many branches of the Federal Government needed to have effective computer communications with the DOD, how would TCP/IP and OSI interoperate? To explore these issues and others, in May 1980, NBS held a workshop entitled: Trends and Applications 1980: Computer Network Protocols.
Meanwhile, DARPA pushed on, and by August 1980, had implemented the new TCP/IP protocol in 12 gateways interconnecting 10 networks.25 In an auspicious sign of the desire to cooperate, in December 1980, the DOD publicly unveiled TCP/IP at the NBS.
The efforts of Xerox’s Systems Development Division (SDD) to upgrade the Pup protocol to XNS progressed nearly in parallel with TCP/IP during these same years. Yogen Dalal, who had assumed management of most of SDD’s networking group after Metcalfe’s departure in late 1978, stayed abreast of changes to TCP and OSI. Dalal remembers:
We were aware of the concept of the Open Systems Interconnection Reference Model. Hubert Zimmerman and those people were trying to define that. We tried to influence many of those thoughts. I guess we were a little bit more applied, so while the Reference Model had a lot of formal verbiage, we tried to inject the layman’s translation of that or what a programmer would think. We considered it somewhat academic, that it was a reference model that attempted to formalize what all of us knew, and the Reference Model, as most models did, attempted to concentrate on the bottom layers because people had experience with the bottom layers, and the higher layers became sort of fuzzy because you hadn’t got to that yet. The influence was nominal, primarily because Xerox made it difficult for us to talk about some of our experiences, and many of our differences on how we viewed the model had to do with the higher level protocols.
In 1979, when Dalal learned from his boss, David Liddle, that Xerox, DEC and Intel were cooperating to standardize Ethernet, Dalal argued that Xerox should also commercialize XNS or try to make it a standard. Dalal remembers asking Liddle:
‘Why can’t we get Xerox to sell or to standardize its network architecture product in addition to its commercial product?’ We had something, we knew it was the cat’s meow, an important revolution, and we were going to make it succeed. Other parts of the company had similar visions of grandeur, but were afraid that if they gave the standard out, they would lose control. That was the biggest mistake Xerox ever made.
Xerox management feared making XNS public even more than Ethernet, for the differentiating intelligence making possible sophisticated printers, print servers and file servers resided in XNS, not Ethernet. The first commercial product built around the new capabilities of XNS and Ethernet was the Xerox 8000 Network System (X 8000) announced in November 1980, just two months after DIX made public the Blue Book and one month before the public unveiling of TCP/IP.26 The X 8000 consisted of a graphics-based user computer (the 860 workstation), laser printers, and various other peripherals, all connected using Ethernet and XNS. Metcalfe, now with 3Com, opines in the December issue of Data Communications that the architecture of the X 8000:
points the way for future design of shared-resource networks.27
- [24]:
Jon Postel,”DOD Standards Internet Protocol,” IEN 128, RFC 760, USC/Information Sciences Institute, NTIS document number ADA079730, January 1980
- [25]:
Jon Postel,”DOD Standards Transmission Control Protocol,” IEN 129, RFC 761, USC/Information Sciences Institute, NTIS document number ADA082609, January 1980
- [26]:
Postel et. al “The ARPA Internet Protocol,” Computer Networks, 1981
- [27]:
Data Communications, Dec 80, p. 26