Chapter 14 - Internetworking: Emergence 1985-1988
14.6 The Role of the National Bureau of Standards (NBS)
In January 1979, the NBS hired John Heafner to manage a newly created Systems and Network Architecture Division (SNAD) of the Institute for Computer Sciences and Technology (ICST) of NBS. All Federal Government local area networking and OSI standards responsibility would thereafter reside in SNAD. Heafner explains the action of the U. S. Congress:
NBS has a mandate from Congress to provide standards for ADP {automated data processing}, communications and so forth. It’s part of their role. The program was reviewed here at NBS in the mid-to-late ’70s, as to why they hadn’t done anything significant, and the answer coming out of that was: “Yes, they haven’t done anything significant, and the reason is they don’t have enough funding.” In 1978, the budget was trebled.
Heafner was well prepared for his new assignment. In 1969, he was responsible for connecting RAND to the Arpanet. When the ARPA sponsored research group at RAND, led by the visionary Keith Uncapher, spun out in 1972 to become the Information Sciences Institute (ISI) affiliated with the University of Southern California, Heafner joined ISI. He then worked at ISI until he moved to the NBS in 1979. Thus he was well aware of many of the dynamics impacting the strategy he would need to fashion, including:
- The obvious fact that the Federal Government was going to buy a lot of computers and increasingly want those computers to share information;
- IBM, the largest vendor of computers to the Federal Government, had zealously protected its proprietary software, including its communication protocols that did not have a Network Protocol like TCP/IP or the planned protocols of OSI; and
- The uncertain outcomes of the considerable investment and knowledge the DoD had in the TCP/IP protocols.
Heafner remembers:
Sort of the way I put it when I came here is that we will have won when DoD buys OSI from IBM. We will have covered all the bases, getting the largest supplier to supply to the largest consumer, and do it in a standardized way. Our thrust from the beginning was to work in ISO, and to some extent in CCITT, to make sure that government needs were covered when those protocols were being developed.
During 1982, NBS held meetings with companies to learn how to move OSI forward, including the use of the recently formulated Manufacturing Automation Protocols (MAP).22 MAP was a defined subset of OSI protocols championed by General Motors. It was clear that a public demonstration of the OSI protocols might force companies to commit to OSI, produce OSI products, and assure their products all worked together, somehow. At the first Implementors Workshop meeting in February 1983 at the facilities of the NBS, the date was set for a public demonstration: the National Computer Conference (NCC) in June 1984.
Heafner aired a proposal one evening at NCC to solve a problem vexing vendors: products that would not interoperate. He recalls:
I presented a white paper proposing the OSINET to the vendors participating in the NCC ‘84 in Las Vegas. At that time, it was called the NBS-CatNet, or concatenated network. It was just several pages of notes, proposing: “Do you want to continue this work?” Vendors do not want to key up for a demo every year or two or whatever. What we needed was some sandbox where we could really play and do inter-operation testing, and develop the test technology and methodology behind all of this. What was envisioned was a sandbox to play in, and that’s what the OSINET is today. I credit Maris Graube with that idea.
By the end of year 1984, the NBS finalized its first Federal Information Processing Standard (FIPS 107) titled: Local Area Networks: Baseband Carrier Sense Multiple Access with Collision Detection Access Method and Physical Layer Specifications and Link Layer Protocol.23 Considerable savings in government expenses were expected, as well as fewer future network integration problems, e.g., instead of every agency continuing to purchase LANs that would some day be unable to be interconnected with other LANs.24 Rosenthal, who had participated in the process from the very first IEEE 802 meeting in February 1980, remembers with pride in 1988:
This is what I make: a Federal Information Processing Standard. It’s the only one that we have in networking in the government right now: Dated 1984, October 31st, and titled: Local Area Networks: Baseband Carrier Sense Multiple Access with Collision Detection Access Method and Physical Layer Specifications and Link Layer Protocol – a mandatory procurement and standard. I told IEEE to let the world know, and put the fact on their cover for 802.3 and portions of 802.2. That’s what this does. So I won at the game. I made the rules, played the game, and I won.25
The challenge for the twenty-one companies that had committed to demonstrating OSI protocols at Autofact ‘85 and, therefore of winning, was to create products that could pass the conformance tests of OSI standards. As it would turn out, conformance testing was a very tall order. Even at that, conformance did not mean interoperability. And there might not have been any tests at all if not for the cooperation between NBS and the participating companies, i.e. the private sector. Mulvenna explains:
This is an area of government/industry cooperation. When we developed the ‘internet’ protocol for the test system for the Autofact, we couldn’t do it. We didn’t have all the in-house resources to do it. Three companies, Intel, Honeywell and NCR, sent us people to work here full time at NBS for four or five months, to develop the test system that we needed to test out the ‘internet’ protocol at Autofact.
- [22]:
See Chapter 11.10 ISO/OSI (Open Systems Interconnection) 1982-1983. Rosenthal remembers: “Now, here’s what happened. In order to pull this off, we had to get some agreements. That’s the key word, ‘agreements.’ We had to get the people highest up in these organizations to commit resources. We had to get a commitment of the CEOs, somebody with signature authority, had to be able to say: “Here’s the check, you make it happen. Pull out all the stops. OSI is important. Make it happen.” We had to get the technical people to ask the question: “Make what happen?” We had to say: “Make this happen,” and we had to lay it out for them. They came back to us immediately and said: “There’s too many options here.” A DEC guy says: “I’m going to implement it this way, and the IBM guy’s going to implement it that way, and we’re at different altitudes. Why don’t we go to the National Bureau of Standards, where we can all be peaceful, sit around the table, our techie guys can sit around and talk to each other without all the fears that you might have,” non-disclosure and all that stuff, we can deal with that here at NBS, “let’s ask the Bureau of Standards, good solid government agency with a good, strong reputation, if they’ll do that for us.” So industry came to the Bureau of Standards and said: “Please help us. Put together a workshop that looks like a big umbrella where out techies can get together, roll up their sleeves, and agree on what it is we’ll implement.” And we said: “That’s a good idea.”
- [23]:
See FIPS 107
- [24]:
In Rosenthal’s vernacular: “I get calls all the time from GSA, The Government Services Administration, saying: “What is this FIPS 107?” And I say: “It’s this Technology,” and people say : “Great.” Anyone in the government can buy products without all the paperwork and shenanigans. They say: “Buy something that FIPS 107,” and it makes life very easy for ADP managers all over the government.
- [25]:
Rosenthal remembers: “Industry can go fight your ring and bus battles until you’re blue in the face. We do not care anymore. We got what we wanted out of it. Let us move on to a richer set of problems: What’s happening at the upper layers of OSI?”