Doug Gale
Recollections, 1980 - 1984
I don't remember all the exact dates, but I do remember the last interview I had with Doug Van Houweling. I thought his timing was impeccable because we reached the subject of salary just as we reached the top floor of the art museum that looked out over Cayuga Lake. It was a beautiful day, the trees were changing color, the blue sky was filled with puffy white clouds, Cayuga Lake was in the background. At the time I remember commenting to him that he had impeccable timing and he said that he worked very hard on that aspect of the recruiting process. At any rate, I agreed to come.
The family and I moved from Texas to Ithaca in December. Our journey was tedious. One day, we'd go "x" miles, the next day "x/2" and the next day half that again. We kept running into problems. We had a trailer hitch that failed while on the freeway. One kid got sick with an earache and the other with something we weren't sure about. As I recall we stopped at an emergency room in the Baltimore area. We finally made it to Ithaca---two days before Christmas. As I recall, Doug Van Houweling and a couple of other folks helped us unload the truck, which included a small TR-4 sports car. We stacked everything in the living room and then we decided, "What the Hell!" and drove back to Kansas City to spend Christmas with family. We got back to Ithaca around the 27th and found some of our Texas friends had already arrived for the first and last annual Ithaca Cross-Country ski event. Since they'd been the first people to stay in the house they took the liberty of arranging the furniture as they thought it should be arranged which took us several years to correct. However, they thoughtfully had unpacked us. As you may recall, that year the Olympics were held in upstate New York and there was no snow short of the Arctic Circle until late in the winter. So we didn't do much skiing. That was my introduction to Ithaca - a cold but very snowless environment with everyone standing around with cross-country skis.
The Terak era: At Cornell, my job as I understood it, was to start a guerrilla movement to change the culture at CCS. We were supposed have so much fun with distributed computing that everybody would look at it and say, "Wow, this is too much fun to ignore." That's basically how it played out. One of my first projects was dealing with the Terak computer labs. Teraks were desktop computers based on the PDP LSI-11 chip from DEC. They were used for a variety of things but the primary user group at Cornell was Tim Teitelbaum's introductory Computer Science course, which was required for all Engineering and Computer Science students at Cornell at the time. There were about 500 students per semester who went through the course. The interesting thing about how Tim taught the course was that the students had to program in PL/1 but they didn't run in a real PL/1 environment. Tim had created a PL/1 synthesizer, which effectively put the students in a padded cell where they couldn't do any damage and where they were also forced to write very "pretty" structured code. They couldn't write ugly code since the synthesizer simply wouldn't let them. If they tried to do something that was quick and dirty and ugly, it just wouldn't work. It forced them to write very good code, which was the unique thing about the system. The challenge was that there were about 500 students each semester in that course, all working in the Terak labs, which had a limited number of machines. There was a lot of contention for the machines and this led to problems. Machines always broke down the night before a major assignment was due. This was two years before the introduction of the IBM PC and 3 years before the Macintosh so there really wasn't much out there in terms of repair facilities. The Teraks were a very limited production DEC system on a desktop and cost $6000 each.
We wound up sending one of our technicians, Tim, to Phoenix, Arizona, and the headquarters of the company that made Teraks. He spent a month there repairing Teraks. They didn't have a repair manual and the technicians just learned by experience. When he got back, we thought it would be a good idea for the other techs to gain some of that experience. One Saturday afternoon with a Camcorder borrowed from faculty member in Uris Hall and a case of beer we filmed "How to Fix a Terak." As you might imagine, towards the end of the day there were lots of cuts in the film as people were rolling on the floor with laughter as the jokes got worse and worse. It wasn't meant to be a professional production but it enabled the other techs to start repairing Teraks. At that time the only alternative was to send a broken machine back to Arizona. That took a day to pack them up, one to two days to get there by air-freight, a week or so to get fixed and then one to two days to get them back. That was pretty excessive and expensive downtime. It was rather interesting that we started getting requests for our video from other institutions that had Teraks. Our rather amateurish production wound up becoming "The" Terak repair tape. It got broader dissemination than we ever anticipated.
We had some interesting experiences in the Terak labs. One of the problems we had was that the facility in the Engineering building, Upson Hall, was quite warm. The students complained bitterly, even during the winter with the windows open, that the room was too hot. In the summer the room was unbearably warm. Our problem was that the maintenance on the Teraks was increasing exponentially---they were failing at an alarming rate. We argued that it was temperature related but Dean Everhart (Dean of Engineering) had a hard time accepting that. He believed that these machines were indestructible and yes, the students were uncomfortable, but it was part of becoming a good engineer. We argued about this frequently and finally in a fit of desperation we mounted thermometers on the top of the CPU chip in one of the Teraks and kept an hour-by-hour running monitor of the temperature of the CPU chip. As I recall we were getting temperatures on the surface on the chip around 150 to 160 degrees. We then took charts of the CPU temperatures to Dean Everhart to try to convince him that we needed more than just a few fans in the windows. We needed serious air conditioning or we weren't going to have any hardware left. We didn't get air conditioning but we did get a few more fans. Unfortunately the Teraks continued to fail as a result of the heat. Of course, the students weren't factored into the equation anywhere but we had some nice charts of failure rates for the electronic devices in warm environments.
The other interesting problems we had with the Teraks was in our own unit, DACS, Decentralized Academic Computing Services. (The name was later changed to DCS, Decentralized Computing Services as we started doing some administrative computing.) One of our early projects (failures might be more accurate) came from trying to provide students access in a fair and equitable way to the Teraks in the open labs. We initially actually used the number of fights which broke out per hour in the queue for the Teraks as a rough measure of access contention. Two fights/hour was crowded, while five fights/hour was very crowded. If there were no fights we concluded that contention wasn't particularly severe even though the students might not be happy. We found the metric distasteful and tried to come up with a scheme that would equitably allocate the machines. The heart of the problem was you'd find some bully monopolize a machine while somebody else was waiting to work on their assignment.
We wanted some way of providing equitable access short of having armed guards in the Terak labs. The solution we came up with was a scheme for using an Apple II computer as a scheduler to control the Terak computers in the labs. As a programming language on the Apple II we selected a language that had been developed in Europe for device control and became obsolete so quickly that I can't even remember its name. There was a small control box beside each Terak, with a wire that went inside the Terak, which would interrupt the system, actually stop it, and cause it to reboot upon command. When the signal came, the job that was being executed was canceled, the machine was shut down, and everything was lost - it was a hard shutdown. There were three lights on the control box. A green light meant everything was cool, an orange light meant you had five minutes before the machine would shut down and then when the light went red the machine was shut down. The control box was controlled by the Apple II, which functioned as a scheduler. Students would come into the lab and rather than forming a physical line as they had previously, they would log on to the Apple II using their user name and password. Their name and place in the queue would be displayed on a big screen along with the estimated time until they would be assigned a machine. A student was limited to one hour on a machine if there was anyone waiting. So far, so good. Students would come in, they would sign in, and wait for a machine to become available. One design problem was maintaining the password directory because the Apple II didn't have much memory. What we did was assign students passwords that were really just a hashed version of their name. The Apple didn't maintain a list of users; it just checked to see that the password entered was the same as the one generated by the name that was entered. This secret never got out and allowed us to use a computer with 8K of memory to monitor a large number of students.
Students would come in, sign in with their name and password, and could see on a monitor roughly how long it would be before they would be assigned a machine. In the meantime other students would be logged off machines. If they ignored the warning, which they would typically do only once, they'd be angry as they realized they had just lost an hour of work. The control machine was absolutely remorseless and absolutely impartial. The students accepted that pretty well. When a student was forced off a machine, the next student would be told that the machine was free. Everything was done in one-hour blocks of time. There was a lot of competition for the machines because Tim gave very, very time intensive assignments and people would tend to wait until two or three days before an assignment was due. Even though the Lab would be open all night, students would be backed up, frantically trying to get their programming assignment done. Everything was cool except there was some kind of very subtle bug in the software which would only show up when the Apple II scheduling machine was very heavily loaded, like the night before an assignment was due. Then the scheduling machine would unpredictably go south and lock up every machine in the Lab. This would only happen in the wee hours of the morning, the night before an assignment was due. It took us a couple of near riots before we acknowledged failure.
Another problem that we had hoped to solve with the Apple logon machine was the lack of utilization statistics. As we talked with faculty and deans around campus who were funding the Terak lab, all of them argued that it wasn't their students who were using the facilities. So as we inventoried the deans, to try to find support for buying more machines we suddenly found that none of their students were using the machines and in aggregate no students were using the machines.
While it seems obvious in retrospect, the ultimate solution wasn't to control or restrict access; it was to convince the Deans that they needed to buy a few more machines.
Macintosh: The Terak era lasted from 1980 to 1984. In 1984 Tim Teitelbaum became very interested in using the new Macintosh Computer and became a factor in the products development. Prior to the public introduction of the Macintosh we arranged a non-disclosure from Apple. Dan'l Lewin and Stacy Bressler showed up on campus with a prototype unit that was basically a box with wires hanging out. They actually bought a seat on the flight to Ithaca so of the prototype would never leave their sight. Dan'l reminded us of a diplomat handcuffed to a briefcase containing nuclear secrets. That's how he treated the prototype Mac. When he arrived at the campus room where we were to have the non-disclosure he just about went into cardiac arrest because there were about a dozen faculty in the room. Outside of Apple, that was more people than had ever even seen a Macintosh. We finally convinced him that these were trustworthy people who would not disclose any of his secrets and he went ahead with the non-disclosure although he did have someone guarding the door for the entire time.
The "gotcha" with adopting the Mac was that Steve Jobs firmly insisted that BASIC was the only programming language needed. The folks in the then forming Apple Consortium of course disagreed and said no - BASIC was for savages! This argument raged back and forth for most of the spring until finally the universities collectively told Apple that if BASIC was all they were going to provide, the universities weren't interested. Jobs finally reluctantly agreed to put Pascal on the machine, which of course was what Tim Teitelbaum wanted, because he was switching to Pascal for teaching his introductory course. Shortly after I left Cornell they put in a laboratory of Macintoshes that were substantially cheaper and easier to maintain that the Teraks.
Following Doug Van Houweling's departure in 1981, Alec Grimison became acting Director of Academic Computing. One day he came into my office and said "I don't want to be Director of Academic Computing anymore, I want to work in DCS and do some fun stuff!" We had a long argument about it and he won and I agreed to do the academic computing stint while he worked on some fun stuff, like working with the Floating Point Array processors. We started a search, which ultimately resulted in bringing in Gordon Galloway as Director of Academic Computing. I then returned to DCS.
About Ken King. I remember that Alec and I had a little book in which we recorded "The Sayings of Chairman King," an obvious reference to another little red book. I've since unfortunately lost the manuscript. However, I remember and continue to use many of them. Friday afternoon Wassails come to mind when I think of Ken King. I remember how he masterfully used the Wassails. He would talk to everybody, encouraging them to have a drink, but if you looked at the amount of alcohol which Ken himself consumed it was virtually zero. He was using the Wassail as a mechanism for getting other people to say what they really thought. It was a masterful stroke.
When I hired Tom Hughes DCS was in crisis mode because the microcomputer revolution was in full force. We had actually instituted "banker's hours" and the door to the DCS office, which was then on the 4th floor of Uris, was locked until noon. At noon we would open up the door and people would come pouring in. It was not uncommon to see eight or ten faculty members waiting in line at Tom Hughes' office. It wasnąt popular but weąd instituted the policy to give our people some time to get work done other than answering peoples' questions. We'd hired Tom fairly early in the revolution. He was very interested in computers. He came from a public school environment where he'd been teaching in a high school and doing absolutely great things. He had a minicomputer on site and was teaching his students programming. This was back in the 70s when this was not the norm. He received absolutely no support from his high school administration. The principal regarded computers as unimportant as everyone knew they werenąt of any importance in education! In 1984 I was amused that at an education seminar in upstate New York his ex-principal gave a presentation on the importance of computers in education. I had a great deal of difficulty in not laughing out loud because they had had a tremendous person on their staff, years before, who had been driven off because the principle had taken such an anti-technology position.
About other staff. One of our first hires was a young woman named Karen Freedman who later went to Computer Sciences Corporation. Of course, there was Alison Brown who had been the previous director of DACS. She had decided that she didn't want to put up with the administrative nonsense and basically recruited me as her replacement. The older I get, the more I understand and sympathize with her sentiment. Anyone who knows Alison, knows what a free spirit she is. She put in tremendous hours and efforts on things but if she decided something was not the right thing to do, she just didn't do it. You managed Alison like you manage cats. It's an oxymoron. You basically let her do what she wanted to do and she did it very well. A later hire was Pete Siegel who became part of the Cornell supercomputing initiative with the deployment of array processors in scientific research. Later Pete headed Cornell National Supercomputing Facility, was executive Director of the Cornell Theory Center, home of the supercomputing facility, and in time became the CIO at the University of Illinois.
One of the downsides of distributed computing was we lost the ability to exchange information that was intrinsic to timesharing. With distributed computing we created islands of small computers. As a result DCS was heavily into high-end computing, Alec's supercomputing efforts, and networking to connect these islands. DCS offered the first general access Unix machine, a DEC minicomputer, on campus around 1983. Alison and I arranged a party and invited Ken King. The party was at the Wilson's house and after we served lots and lots of spaghetti and cheap red wine, we rolled out our proposal to bring in a university UNIX system. When the system arrived we found that there wasn't any networking software and it wouldn't support an Ethernet connection. So Alison wrote the networking drivers for that early PDP-11 while she was at DCS. As we were getting into the networking stuff, Alison took the lead. Her strong interest in networking was one of the reasons she took a position with OARnet and the Ohio Supercomputing Center.
After Cornell: I left Cornell in the summer of 1984 and went to the University of Nebraska to become their Director of Computing. The primary reason for the move was family. My mother, who lived four hours away in Kansas City, was entering her 80s and her health was declining. At the time, we only expected her to live a year or two. I didn't appreciate what a tough old bird she was. She lived another eight years. At any rate, that was the primary reason for leaving.
In 1985, while I was at UNL, I wrote a one-page unsolicited proposal to the NSF that suggested connecting the supercomputing centers with 56KB lines as a backbone network and in addition give me some money to create a regional network between the Big 8 institutions in the Midwest. Interestingly enough, Henry Schaefer at North Carolina and Glenn Reichert at Maryland wrote a similar one-page letter. We didn't know each other well at the time. None of us heard anything for a year but in the summer of 1986, as we put this together later on, we got phone calls from NSF that asked if we remembered the proposals we had sent to them. Our somewhat sarcastic answers were "Vaguely!" They asked us to flesh out the proposals and send them back. My response was, "That's great, how about next week!" The response I got back was, "How about yesterday!" So we compromised on tomorrow. I stayed up all night and fleshed it out. That was the birth of the NSFNET and MIDNET in the mid-west, and SURANET in the southeast. NYSERNET, started by King, Mandelbaum and Schrader, started about two weeks later. Those three networks where the first networks on the NSFNET. Ever since then we have had an argument about who was the first to go live. Each of us was first depending on how you define "live".
While at UNL I was on leave for two years, 1990 and 1991, to serve as the Director of NSFNET at the National Science Foundation. I stayed 11 years in Nebraska before going to George Washington University as their CIO. In 1998 I took a position as the head of OARnet, which ironically Alison Brown had founded after she left Cornell. In 2002 I left OARnet, set up a small consulting firm and moved to the mountains of Big Sky Montana.
Prepared on October 27, 1999 by John W. Rudan