The Beginnings of Computer Networks
4.2 Host-to-Host Software: The Network Working Group 1968-1969
Roberts never questioned that each site would have to connect its computer to the subnet. For to convince sites that an outside contractor would be responsible for software residing on “their” computers was a thought so top-down as to undo the very heart of ARPA logic, that is, hire the best and leave them alone. He needed the Arpanet to be seen as the users’ creation, not a straightjacket imposed from on high. Thus the Host sites had to create the architecture and software that would make the Arpanet a functioning network.
So concurrent with the release of the RFQ in June 1968, Roberts asked Elmer Shapiro of SRI to organize representatives from the initial four sites to begin working on the host-to-host issues. Shapiro chaired the first meeting of site representatives in early July 1968. This meeting began a process that would become even more problematic than the subnet development. Steve Crocker, who attended as the representative of UCLA, remembers:
"The first meeting was seminal. We had lots of questions -- how IMP's and hosts would be connected, what hosts would say to each other, and what applications would be supported. No one had any answers, but the prospects seemed exciting. We found ourselves imagining all kinds of possibilities -- interactive graphics, cooperating processes, automatic data base query, electronic mail -- but no one knew where to begin. We weren't sure whether there was really someone from the east that would be along by and by to bring the word. But we did come to one conclusion: We ought to meet again."
Around the same time, Taylor initiated an annual meeting of graduate students working on ARPA contracts, a meeting similar to the annual principal investigators conference. The first meeting was held in July 1968 at the University of Illinois, shortly after the Shapiro meeting, and was chaired by Barry Wessler, Roberts’ assistant since early 1968. Crocker attended as the MIT representative, and Alan Kay and John Warnock - not Steve Carr - showed up from Utah. Vinton (Vint) Cerf, a high school buddy of Crocker's who had worked for IBM and was now in the Ph.D. program at UCLA, represented UCLA. Crocker recalls:
"Barry Wessler kept trying to get us interested as a group in this network, and the group's reaction was not very positive. It wasn't negative. There was no protest, but he wanted us to think about the issues of file sharing and some of those kinds of issues. We were all interested in artificial intelligence and graphics systems and other topics that were hot at the time, and there wasn't a lot of response. It just sort of fell flat. So, the one thing that was true was that the community was not keenly involved in networking, and not really responsive. On the other hand, ARPA was pushing it and it was a serious vision."
During the summer and fall of 1968, Crocker, Carr and Jeff Rulifson began meeting as Arpanet site representatives. Crocker remembers:
"We didn't have a charter. There wasn't any agenda. There wasn't any mandate to go forward. There wasn't any organization imposed. There wasn't even any authority to do anything. We had been told, basically, that these IMP's were coming and we would get connected, and that was all the structure there was.”
In February 1969, Heart convened a meeting of site representatives at BBN to begin the process of educating the sites about the IMPs, so they would be ready to connect to them when they were installed later in the year. Crocker, Carr and Rulifson attended hoping to be told "what to do." Crocker remembers:
" I don't think any of us were prepared for that meeting. The BBN folks, led by Frank Heart, Bob Kahn, Severo Ornstein and Will Crowther, found themselves talking to a crew of graduate students they hadn't anticipated. And we found ourselves talking to people whose first concern was how to get bits to flow quickly and reliably but hadn't -- of course -- spent any time considering above the link level."
BBN concluded that they had to provide information to the host sites if they were to develop the hardware and software needed to interface to the IMPs. Kahn assumed responsibility for creating a document specifying the IMP interface, a document to become BBN Document #1822.
Thanks to the February meeting, Crocker, Carr and Rulifson realized that they were the leaders. If the sites were going to connect to the Subnet, they had to make some decisions and get to work. At their March meeting, they decided to begin documenting their conversations, not as formal minutes but as notes, a trail of thoughts, to which they gave the name Request for Comments. Crocker, who had become de-facto chairman of the small group of site representatives, submitted the first Request For Comment: RFC 1 - Host Software, on April 7, 1969.
In RFC-1, Crocker described how the IMP software worked and its implications for Host software. Significantly, he described that when a Host wanted to initiate a connection with another Host, it did so by sending a link code that the IMP would use to establish a link with the intended Host. A second message could not be sent over an established link without receiving an RFNM (Request for Next Message). Each Host would have only so many links, and would have a limited number of links with each other Host. The creation of connections using links was tantamount to establishing virtual circuits between Hosts. Hence, while the subnet functioned by routing packets, Host connections functioned by sending messages over virtual circuits.  This distinction and its implications would embroil the computer communication community for over a decade and is, therefore, central to this history.
Another point Crocker made in the RFC was the desire for some host-to-host error checking. BBN made it clear that no checking was needed, as the subnet would provide sufficient error-correction. This assumption would prove insufficient and, combined with the fact that no host-to-host checking would be built-in, caused future problems.
In June 1969, BBN distributed Document #1822 to the sites and their representatives. It defined the physical, link level and packet level interfaces to the IMPs. With only three months to go before the sites began receiving their IMPs, site personnel needed help. Heart recalls:
“BBN put a great deal of energy into helping the host sites. We not only wrote 1822, but we didn't just go and leave it. We went and involved ourselves with every single site in those days, talked to the people designing the hardware and software, helped them over rough spots. We put a lot of energy into working on the host side of it. Even though we couldn't do it ourselves, we were very heavily involved. So for us, the problem was getting those hosts on, and we wanted that to work, otherwise the network wouldn't be used and it wouldn't grow. It was an important issue, and of course ARPA also was very anxious to make it happen, and encouraged us to do that.”
Technically, the subnet also created circuits and was not truly a packet switching.