Crocker was among a small group of UCLA researchers who sent the first message between the first two nodes of the ARPAnet, the U.S. Department of Defense–funded network that eventually morphed into the modern Internet.
Crocker’s biggest contribution to the project was the creation of the Request for Comments, or RFC. Shared among the various research institutions building the ARPAnet, these were documents that sought to describe how this massive network would work, and they were essential to its evolution — so essential, they’re still used today.
Like the RFCs, Crocker is still a vital part of the modern Internet. He’s the chairman of the board of ICANN, the organization which operates the Internet’s domain naming system, following in the footsteps of his old high school and UCLA buddy Vint Cerf. And like Cerf, Crocker is part of the inaugural class inducted into the Internet Society‘s (ISOC) Hall of Fame.
This week, he spoke with Wired about the first Internet transmission, the creation of the RFCs, and their place in history. ‘RFC’ is now included in the Oxford English Dictionary. And so is Steve Crocker.
Wired: Some say the Internet was born on Oct. 29, 1969, when the first message was sent between UCLA and the Stanford Research Institute (SRI). But others say it actually arrived a few weeks earlier, when UCLA set up its ARPAnet machines. You were there. Which is it?
Steve Crocker: October. The very first attempt to get some communication between our machine, a Sigma 7, and [Douglas] Engelbart‘s machine, an SDS-940, at SRI.
Famously, it crashed.
We tried to log in [to the SRI machine]. We had a very simple terminal protocol so that you could act like you were a terminal at our end and log in to their machine. But the software had a small bug in it. We sent the ‘l’ and the ‘o,’ but the ‘g’ caused a crash.
Their system had the sophistication that if you started typing a command and you got to the point where there was no other possibility, it would finish the command for you. So when you typed ‘l-o-g,’ it would respond with the full word: ‘l-o-g-i-n.’ But the software that we had ginned up wasn’t expecting more than one character to ever come back. The ‘l’ was typed, and we got an ‘l’ back. The ‘o’ was typed, and we got an ‘o’ back. But the ‘g’ was typed, and it wasn’t expecting the ‘g-i-n.’ A simple problem. Easily fixed.
Wired: And the Internet was born?
Crocker: Some say that this was a single network and therefore not ‘the Internet.’ The ARPAnet was all one kind of router, and it didn’t interconnect with other networks. Some people say that the Internet was created when multiple networks were connected to each other — that the IP [Internet protocol] and TCP [transmission control protocol] work on top of that were instrumental in creating the Internet.
The people who worked at that layer, particularly Vint Cerf and Bob Kahn [the inventors of IP and TCP], tend to make a careful distinction between the APRAnet and the later expansion into multiple networks, and they mark the birth of the Internet from that later point.
But, conversely, the basic design of protocol layers and documentation and much of the upper structure was done as part of the ARPAnet and continued without much modification as the Internet came into being. So, from the user point of view, Telnet, FTP, and e-mail and so forth were all born early on, on the ARPAnet, and from that point of view, the expansion to the Internet was close to seamless. You can mark the birth of Internet back to the ARPAnet.
Wired: It was before that first ARPAnet transmission that you started the Requests for Comments. They helped make that transmission possible?
Crocker: The people at ARPA [the Department of Defense’s Advanced Research Projects Agency, later called DARPA] had a formal contract with Bolt, Beranek, and Newman [or BBN, a Boston-based government contractor] for the creation of the routers, and they had a formal contract with AT&T for the leased lines that would carry the bits between the routers, across the country. But they had no formal plan, or formal paper work, for the nodes that would be connected to the network.
What they had instead was a captive set of research operations that they were already funding. The first four [nodes on the ARPAnet: UCLA, SRI, University of California, Santa Barbara, and the University of Utah] and all of the other places that would play a part in those early days were places that were already doing research with ARPA money.
These were pre-existing projects of one sort of another. Graphics. Artificial intelligence. Machine architectures. Big database machines. All the key topics of the day. Douglas Engelbart’s work at SRI was focused on human-machine interaction. He had an early version of a mouse and hypertext working in his laboratory, for example….
So, the heads of each of these projects were busy with their own agendas, and here comes this network — which was kind of foisted on them, in a way. Not unwillingly, but not with any kind of formality either. So, basically, they delegated the attention to this project down to the next level. In the case of the university projects, that meant graduate students, and in the case of SRI, that meant staff members below the principal investigator level.
Somebody called a meeting in August of ’68, and a few of us came from each of these places … on the order of a dozen or fewer people. Vint and I drove up from L.A. to Santa Barbara, where the meeting was held and met our counterparts. And the main thing that happened was that we realized we were asking the same questions and that we had some commonality in our technical backgrounds and our sense of what should be done — but there wasn’t a lot of definition to it.
So we made one of the more important decisions, which was to go visit each other’s laboratories and to keep talking to each other. And we understood the irony that this network was supposed to reduce travel and the first thing we did was increase travel.
Over the next several months, from August ’68 to spring ’69, we had a series of meetings where we visited each other’s labs, and we also had kind of freeform discussions on what we might do with this network — how it might develop. We didn’t have a detailed specification of how the IMPs [interface message processors] were going to be connected to the hosts.
When we started, BBN hadn’t actually been selected. I think they were selected and got started on the first of January 1969. Some us went out to meet them in Boston in the middle of February 1969, in the middle of a large snow storm. But they didn’t publish a detailed specification of how you connect a host to an IMP until later that spring. So [the researchers] had this time from when we first met each other to the time we had a detailed spec in which we could speculate and focus on the larger issues without having to narrow down into ‘this bit has to go here’ and ‘this wire has to go there,’ and we started to sketch out some key ideas.
There was no senior leadership. There were no professors. There was no adult in the room, as it were. We were all more or less in our mid-20s and self-organized. Out of that emerged … a strong sense that we couldn’t nail down everything. We had to be very ginger about what we specified and leave others to build on top of it. So we tried to focus on an architecture that had very thin layers that you could build on top of — or go around.
After a few months, we had a meeting in Salt Lake City, and we said: ‘It’s time to start writing some of these ideas we’ve been kicking around.’ We made assignments to each person, and then I took on the minor job of organizing all these notes.
I found myself very nervous about that over the next couple of weeks. At first, it seemed simple, but then I realized that the mere act of writing down what were talking about could be seen as a presumption of authority and someone was going to come and yell at us — presumably some adult out of the east, either Boston or Washington.
So I got increasingly nervous. I was staying with some friends in the Pacific Palisades area, and late one night, I couldn’t sleep and the only place I could work without waking people up was in the bathroom. It was 3 a.m., and I scribbled down some rules for these notes. I said that they were completely informal, that they didn’t count as publications. You could ask questions without answers. You just had to put your name and the date and a title on these things, and I’d assign them a number as fast as you wrote them.
There wasn’t any editorial control. And then to emphasize the informal nature, I hit upon this silly little idea of calling every one of them a ‘Request for Comments’ — no matter whether it really was a request or how formal or how informal.
I genuinely thought that by the time the network got built in the fall, there would be some formal documentation and that these notes would just become obsolete and be thrown away. But they stuck. And this became the primary mode of documentation. And it persists today, although it has gone through some transformation.
Wired: How close are today’s RFCs to the original version?
Crocker: In the early days, we didn’t have the network. We were anticipating it. So some of the early RFCs were lists of people to send the RFCs to. There were a whole series of RFCs that have nothing more than the mailing list and changes to that mailing list. Obviously, that’s not interesting today, when you have e-mail.
E-mail also changed things in that you don’t have to write a full document to discuss something. You can just send an e-mail to a list.
In the early days, we had a forerunner to the Internet Engineering Task Force, what we called the network working group. After the first several meetings, it grew and grew. It got to be 20 people and then to 50 people, and it got to be so unwieldy that we had to divide it up into two parallel sessions to have different discussions. Today, there are on the order of 100 working groups operating in parallel.
People come from all over the world for the physical meetings, which are three times a year, and they range from 1,200 to 1,500 people. But of course, most of the work is done over the net with e-mail.
Now, there’s also an intermediate form that evolved called Internet drafts. After several iterations of Internet drafts, the working group working on something will say, ‘We’re done with this’ and then they’ll get it approved and published as an RFC.
We passed through 1984 some time ago. So RFC doesn’t mean anything like a Request for Comments anymore. It means a formal publication. Perfectly Orwellian.
Wired: You left UCLA and the ARPAnet in the middle of 1971. But in June of 2011, you came full circle, taking over as the chair of ICANN’s board of directors, like Vint Cerf before you.
Crocker: I spent a long time working on other things — principally formal proof techniques for software, program verification — but gradually, over time, I got more involved with computer security and network security and got back more deeply into the whole Internet culture….
In the ’90s, at the IETF [Internet Engineering Task Force], they created a security area, and I got invited to be the first area director and that put me on the steering group for the IETF. I did that for several years and was later on the Internet Architecture Board….
Later, Vint became the chairman of ICANN, and when 9/11 came along, ICANN, like every other organization in the country, said to itself: ‘Security is really important. What else should we be doing?’ And it formed a security and stability advisory committee, and Vint asked me to chair that, so I took that one. He said six months, but I knew it would be a couple of years minimum, and it dragged on, and I got more deeply enmeshed in ICANN. The committee I was on eventually got a non-voting seat on the board, and I took that role as well. Eventually, I shifted over to a voting position, and longevity just wins out here.
Wired: Is this in any way like the work you first did on ARPAnet more than 40 years ago?
Crocker: Jon Postel was another member of the group at UCLA, and when I left, I sort of turned to him and said: ‘Hey would you take the RFCs on?’… Over a period of time, that little insignificant role of assigning numbers to RFCs morphed into a lot more bookkeeping, all of which fell to John. Then the domain name system was created under his watch, and there was yet more bookkeeping. He would assign top-level domains to various countries.
He was also an active researcher, well respected and accepted in the inner circles. So he had this dual role, being inside the network research community and carrying on this clerical function on the side, and that persisted for an amazingly long time without much modification. He changed jobs a few times and kept those duties with him.
Eventually, it became a little dicey. Different parties and countries said: ‘How come it’s being controlled this way?’ And the University of Southern California, where he was working, got nervous about this, the White House got involved, and ICANN was formed.
He was slated to become the chief technology officer, but he died almost at the moment that ICANN was formed. It was an unfortunate. But there’s an legacy that lives on, that’s rooted through him, all the way back to the earliest days of the ARPAnet. And in a certain sense, a provide a bit of continuity, reaching all the way back to that time.