San Jose
When people ask me to explain how INTEROP is different from other trade shows and conferences, I like to describe it as resembling a circus, but not a zoo. A circus appears to be all chaos, but it is carefully managed chaos. A zoo, on the other hand, puts the animals into cages and lets them do what they want.
Networld (referred to by one wag as “Notworld”) is the trade show equivalent of a zoo. Each little cage has its own independent little show. In some cages, exuberant marketing types reach out to touch you with their brochures. In others, they stand listlessly around. The visitors to Networld stroll from one cage to another, feeding their name tags to the animals, hoping that one of them will do something interesting.
INTEROP, on the other hand, is a show. Granted, there are also cages, but those are a small part of what is going on. There are tutorials, a conference, and, most importantly, a huge, operational network.
When I arrived on Saturday, October 5, 1991, people were just beginning to descend on San Jose. After checking into the luxurious Holiday Inn, I wandered over to the Fairmont Hotel. In a meeting room, about 50 volunteers, engineers from all over Silicon Valley and from around the world, were gathering.
All were quite bleary-eyed. Most of them had been up all night stringing cable in the Diamond Pavilion, one of the auxiliary exhibit halls being used by INTEROP. This gathering was getting ready for the real work that lay ahead that night: installing the network at the San Jose Convention Center.
All of the vendors (with a couple of minor exceptions like my publisher) were connected to a real operational show network, which in turn was connected to the Internet. This was no coaxial cable down the center of an exhibit hall. The network used over 35 miles of cable and connected 300 vendors together with what one observer estimated would be the equivalent of the components needed to wire a 20-story high-tech skyscraper.
The organizing principle for the network was a series of ribs. Each rib ran through a physical area, such as an aisle in the convention center or the Diamond Pavilion. Each rib had an Ethernet based on unshielded twisted pair and a 16-Mbps token ring. Each of the 25 ribs had a 19-inch equipment rack at the end with the electronics to drive the subnets and routers to be connected to the backbones.
Two different backbones connected the 50 subnets, one based on FDDI, the other on Ethernet. A T1 line connected the Diamond Pavilion to the Convention Center. A microwave link went to two terminal clusters nearby in the Fairmont Hotel. Yet another T1 link linked the show network to the NASA-Ames Research Center, which in turn provided links to the Bay Area Research Network and out to the NSFNET backbone.
This entire network was put in by a semifanatical team of volunteers, overseen by two INTEROP staff members frantically trying to keep some sense of order. The challenge to this network is that you don’t get convention centers a month ahead of time to put in your network. In fact, as we met Saturday afternoon, they were still tearing down the booths from the previous week’s Seybold Publishing Conference.
At midnight, the Convention Center would be turned over to the Interop Company. At 8 A.M., just eight hours later, dozens of 18-wheel tractor trailers would roll onto the convention floor to start setting up the vendor exhibits. By that time, all the cable needed to be off the floor: try telling a teamster on a forklift to take the long way around because he might crush the fiber.
Between midnight and 8 A.M., 35 miles of cable had to be rolled out and hung from the ceiling with cherry pickers. Equipment racks had to be moved safely into place, and equipment for the Network Operations Center (NOC) had to be moved up into the control booth overlooking the convention floor.
To make this all happen, a core team of a half-dozen volunteers met with Interop all year to plan the network. In July, the company got the convention center for a day and hosted a cable laying party. All the cable was laid out and connectors added. Then, the cables were tied together and carefully rolled onto drums and moved into a warehouse. Before the show, there was a hotstaging in the warehouse, where the cables were connected to equipment racks to be tested. Then, everything was packed up onto pallets to await the teamsters.
At midnight, a group of about 50 had gathered in the lobby of the convention center. Each member of the core team had a different colored shirt, each with the words “Do Not Disturb” stenciled on the back.
I was assigned to the teal team, under the leadership of Karl Auerbach. Karl was one of the founders of Epilogue Technology and a long-time participant in the INTEROP ShowNet. He’s also a lawyer, which makes him a formidable rabble rouser at IETF meetings.
With a loud bellow, we were all called over for a briefing by Stev Knowles. (The ending “e” in “Steve” got left off of a mail message once, and Stev decided he preferred it that way.) Stev is vice president of engineering at FTP Software and is widely acknowledged as the loudest member of the ShowNet team.
Stev’s briefing was, as usual, direct and to the point. “Do what you’re told and if you have a question, ask.”
Stev is an interesting character. Rumor is that he got involved in the very first ShowNet because he couldn’t read his mail. The network wasn’t working, so he marched into the show and commandeered the ShowNet team until things started working.
By sheer force of will, he and the other core team members do this every year, staying up for several days straight to get the network up and running. For this, everybody gets a t-shirt. Of course, the core team got put up by INTEROP at the Fairmont, but even the impressive bar bills they ran up didn’t quite explain why they did this.
A few minutes after midnight, the doors opened and we all stood on the cavernous convention floor, clustered around our leaders. Two tractor-trailers were driven into the center of the floor and began to dump their contents. Equipment racks, spools of cable, and various other network paraphernalia were all hustled to their proper locations.
The barrels of cable began to be unspooled. Teams of volunteers, spaced every few feet, would march a string of cable across the hall. A few ties holding cable were cut and the feeds that would hang down from the ceiling were separated from the main rib.
Then, five cherry pickers started a slow march down the convention floor. At each rib, the cherry pickers descended, then in unison (or some ragged semblance of unison) lifted the cables to the rafters.
Hanging down at intervals were coils of cable. These feeds were left about 16 feet off the ground, high enough to clear the semis coming in the next morning but low enough to reach without a cherry picker.
By 5:30 A.M., most of the ribs were successfully up, so I went back to the luxurious Holiday Inn for two hours sleep before I started my day of meetings. The volunteers were still working when I left and would continue to do so for the next two days straight. Getting the cables up was the most time-critical task, but plenty of work still remained. Cables had to be connected to equipment racks, connectors tested, the backbone had to be tested, vendors had a million questions, the Internet link needed to be initialized, and a million other details had to be finished before the show opened on Wednesday.
The volunteers that put in this network are an amazing bunch. Some of the best network managers from all over the country come just to help out. When they are done, they have a complex internetwork up and running and connected to the rest of the global mesh. Three days later, it is torn down.
While the ShowNet is operational, it supports some fairly heavy duty applications. Groups of vendors get together to demonstrate the interoperability of standards such as Frame Relay, SMDS, X.400, SNMP, and many others. In marketing-speak, these are called Solutions Showcases. As you walk through the exhibits on the convention floor, different booths have little signs indicating which showcases they are part of.
The Solutions Showcases are certainly useful as a marketing tool. Lots of users come to the show to see what works and who sells the equipment. However, the showcases are equally important to the engineers that design and make the technology.
I spoke to one engineer who says he gets more bugs worked out in one week at INTEROP than he can in six months in the lab. By testing his implementation with those of other vendors, he can quickly hone in on ambiguities in the standards and figure out what to do to make the standard an interoperable reality.
The INTEROP week passed by quickly. The first two days, I sat in on a few of the tutorials, hearing people like Craig Partridge talk about gigabit networks or MIT’s Jeffrey Schiller talk about network security and Kerberos. On Wednesday, the show officially began and the pace picked up. The plenary address was via T1 video link from Geneva. I was particularly interested in this address since I was going to need the same satellite link on Friday.
Things went smoothly Wednesday morning, mostly because the previous two days had been spent in frantic preparation. The video link was donated by Sprint. Coordination of this part of the show at INTEROP was handed to Ole Jacobsen. Ole had worked long hours with Sprint technicians, watching the line go from Geneva to Atlanta until, the day before, it had reached Kansas City. By Tuesday night, video was finally making it all the way from Geneva to San Jose.
Ole was the perfect choice to handle anything telephone related. Officially, he is editor and publisher of the ConneXions journal. Unofficially, he is a phone junkie. For example, he has a PBX in his house. Of course, the PBX is a small one with only 6 lines and 16 extensions, but how many houses do you know with their own PBX? His wife Susan still can’t get used to dialing “9” for an outside line.
After the plenary, I dove into the exhibition and conference. Most people were in high-speed data acquisition mode, trying to work out optimal patterns for navigation of the floor, or flitting from one conference session to another hoping that time division multiplexing would enhance their information intake.
My personal favorite, as it had been the year before, were the SNMP demonstrations. In 1990, John Romkey had developed the Internet toaster. The toaster was hooked up to the ShowNet with a PC running TCP/IP and SNMP software. Workstations around the show floor had network management software, complete with a toaster Management Information Base (MIB). By setting variables on the toaster MIB from a network management station, people could make the toaster start toasting.
The problem everybody immediately saw with the 1990 demonstration was that you had to put the toast into the toaster. This was suboptimal from the point of view of the network engineer, who wanted to stay in bed while breakfast was made.
This year, the FTP Software booth sported the new, enhanced Internet toaster. Using the latest Lego technology, they built a little crane that would pick up a piece of bread and deposit it into the toaster slot.
What made this demonstration especially funny is a bit of an inside joke. If you read Marshall T. Rose’s The Simple Book, you will see that he feels strongly that people should make better use of the get-next operator for efficiently using the network and the remote management agent.
To get his point across, Marshall always refers to the get-next operator as the “powerful get-next operator.” The INTEROP audience was truly impressed that the powerful get-next operator was being used to make toast.
Friday morning was the official unveiling of the activities of the Document Liberation Front. I got up early and went over to the Center for the Performing Arts, the biggest hall at the INTEROP show, seating 2,701 people. This hall is so big that it even has enough room to hold the hundreds of students that show up for the tutorial taught by Doug Comer of Purdue, a researcher well known for his explanations of TCP/IP fundamentals through his tutorials and books.
The members of the panel started to show up. Dr. Vinton Cerf, chairman of the Internet Architecture Board, was there to lend his support to the announcement. Richard desJardins, head of a training group called The GOSIP Institute, was there to provide his commentary on the different models of standards making.
We finally got through to Geneva. Although neither Tony Rutkowski nor Dr. Tarjanne was present yet, we held our breath and got ready to start the session. Looking out into the audience, which contained no more than 50 people who had bothered to get up for the early morning session, I was very glad that the cameras couldn’t beam back to Geneva.
We made our announcement and the one member of the press to show up furiously took notes. The announcement went well, but I was worried that nobody would care and that all this effort had been in vain.
Over the next two months, I was to discover how wrong I was. Leaving INTEROP and traveling around the world, I got more news at every stop on how popular Bruno and the Sons of Bruno had become. I thought the choice of Giordano Bruno (1548–1600) as the namesake for my server in Colorado had been an appropriate one.
The Greeks, in those days before the printing press, used a mnemonic method for remembering verse or other forms of knowledge. The Dominican order had kept alive these secrets during the dark middle ages, but had kept them tightly guarded. Bruno joined the Dominican order in 1565 and mastered the Dominican secrets. Then, he revealed the secrets to the rest of the world in his classic, Shadow of Ideas (1582).
Bruno was expelled from the Dominican order, and was later denounced to the Inquisition in Venice for acts of heresy, including telling jokes in poor taste about God. He was burned at the stake in 1600.
Would Bruno’s secrets help the world? Would the ITU Inquisition kill our server? These and other questions were on my mind as I boarded a plane Sunday, bound for Hawaii and then onwards, three times around the world in the next six months.