AIM INTERVIEWS

Who are you?

Steve Crocker. I was part of the team of students at UCLA that put the first host on the Arpanet and worked on the initial set of protocols for the Arpanet. I designed the Host-Host protocol, later known as the Network Control Protocol (NCP). I provided a tiny bit of administrative glue to the ad hoc Network Working Group (NWG). As part of that process I initiated the Request for Comments (RFC) series of notes. Bill Duvall from SRI gets credit for suggesting the term. At the time, I thought the notes would be temporary -- six months or a year -- to be overtaken by more professional documentation. Charley Kline, Mike Wingfield, Jon Postel and Vint Cerf were also part of our group. Professor Len Kleinrock was the principal investigator.

As a member of the early UCLA team working on ARPANET in 1969, you helped set up the first IMP and send the famous first ARPANET message. What originally made you interested in working on ARPANET? Was there any inspiration?

Heh. The famous message you're referring to was the first attempt to log into the SRI 940 in Doug Engelbart's lab from a terminal connected to our Xerox Sigma 7. The SRI machine had command completion, so if you typed enough of a command for the system to determine that it could be completed in only one way, the system would spit out the remaining characters. Thus, instead of typing LOGIN, only LOG was needed, and the 940 would add IN. Bill Duvall's initial implementation of the software on his end didn't anticipate receiving three characters instead of one. (Space was very tight on those machines, so it was natural to allocate just the minimal amount needed.)

Late one evening when Charley and Bill were ready to test, Charley typed L. It went from the Sigma 7 to the UCLA IMP to the SRI IMP and then into the 940. The 940 duly echoed it back, and it showed up on Charley's terminal. Same with the O. But when Charley typed the G, Bill's software hiccuped because the 940 returned GIN. Easily and quickly fixed and not worth mentioning except Charley carefully documented it in the logbook in the lab. Len read the log the next morning and publicized it widely.

I had a bit of experience with system software and operating system design. I had planned to work on program verification but when the Arpanet project came along, I thought I could help, and I thought it was worth a short detour :)

You've mentioned in the past that many elements of the design for the Network Control Program (NCP) and the creation of the RFC series were largely accidental than carefully planned. In hindsight, what was the most surprising outcome or development that came from these?

The Arpanet was designed to deliver messages within 1/2 second with perfect reliability. We tried to include an end-to-end checksum as a way of isolating possible configuration errors. (We weren't concerned about transmission errors. The IMPs had very strong checksums and retransmission to cover that class of errors.) Frank Heart, the manager of the IMP project at BBN, objected vigorously that we'd make the network look slow. We removed the checksum from our design. As it turned out, there were a couple of problems that occurred later which would have been easier to track down if the checksum had been included. And, of course, everyone learned that even if the transmissions are perfect, congestion could consume all the available storage in the IMPs and cause the entire network to lockup. The only solution is to tolerate dumping packets. When the NCP was replaced by TCP, checksums and retransmission were included.

The RFCs were started while we were designing the protocol. We didn't yet have the network, so they were distributed by snail mail. We imposed only minimal rules in order to encourage unfettered interaction. See RFC 3. Over time, the RFCs became the vehicle for formal documentation of the protocols, and the publication process evolved. Email and Internet Drafts became the way to communicate rapidly and with less formality. (Although RFCs are the way the IETF publishes formal protocol specifications, RFCs are also used to publish other classes of documents.)

You've known fellow internet pioneers and UCLA ARPANET colleagues Vint Cerf and Jon Postel since high school. Would you say there was something unique about your shared interests growing up or the environment at UCLA that led to the three of you becoming key contributors to the early internet?

Vint and I did indeed meet in high school. Jon went to the same high school but we didn't meet him until we were all in the same group at UCLA. In those days, most computers were operated in batch mode. Time-sharing was much more appealing, and the idea of a network of time-shared machines to support interactive research was alluring.

Are there any people or projects you particularly admire today?

I had the pleasure of working for Larry Roberts at (D)ARPA. He was amazing. The key people who founded the AI field -- Herb Simon, Al Newell, Marvin Minsky, John McCarthy -- were other intellectual giants that inspired me.

Vint Cerf has astonished me continuously over the 60 plus years I've known him. Enormous breadth, unbelievable energy, unshakeable temperament and rock solid values.

What's a message you have for the world?

Work on projects that inspire you and will have a positive impact on the world. Take chances, but maintain your physical and mental health. Enjoy life.