Interestingly, the old man seems to like China very much. He has been to China and is still learning Chinese. The stubborn old man left Stanford and Sun and started his own Peerouette company.
Q: Tell me the story of developing the first router when you were a researcher at Stanford.
A: This project started in 1 month, 1980. At that time, the boss said that you are engaged in the internet. Let's see how the computer department, the hospital and the electrical department are connected.
Q: What kind of devices need to be interconnected?
A: We have mainframes, DEC 10 systems, some Xerox PARC Lisp machines, Altos file servers and printers. A few years later, we added DEC VAXs, Texas Instruments Explorers and Symbolic systems. These must be related, because we are scattered all over the campus, and everyone is tired of running around with tapes.
I thought about it and said that what we really needed was an operating system. During the test, the network operating system and routing code were developed on DECPD 1 1/05. But Allen Snyder's portable C compiler produced a lot of code. I want to study compiler optimization code generator, but it is not good enough. So I had to rewrite an optimized compiler for PDP 1 1/05, which can reduce the code by about 30%. This is a big project, because everything needs you to do. You should know that PDP 1 1/05 only has 56KB of user memory, and it is diskless. Too many restrictions, you have to do a lot of things in memory allocation algorithm, memory overflow and so on. I spent the whole summer to ensure that the scheduling and packet switching algorithms of the network operating system are optimal.
It took three months to get the basic prototype of the system, and another six months to place the first router in the telephone room of the pine hall. PineHall is located between the hospital and the computer science department, and the cables on both sides of the router are 2000 feet long.
Q: What protocols does this router support?
A: First of all, the device routes Parc general data packets between the Xerox PARC system and the host. Then 8 1, it is said that IP appeared. See what you can do, and then I got some IP routing support. At that time, you don't have to consider ARP, because it is 3M Ethernet, and your IP address is 2 bytes. The first byte is your network address and the second byte is your host address. By 1983, the router has been able to support XNS (Xerox Network Service), TIEExplorer between Chaos and ymbolics Lisp machines and IP. That's when Stanford started using 10M Ethernet.
Q: Did you replace the router based on PDP 1 1/05 with the 68000 motherboard platform developed by Andy Bechtolsheim?
A: Andy is a master student, and the opportunity is very accidental. At that time, we heard about his board, and after talking to him, he agreed that we could have one. We plugged it into the multi-bus backplane, plugged in some 3Com Ethernet boards, and then made several copies, and then I started transcoding. The limit at that time was the speed of the bus. Bechtolsheim's board has 256KB of memory, which is heaven for me.
Q: Is that when the campus network started?
A: At first people were a little skeptical, but it was not until 1983 that it was clear that this was king. At first, it was just technical participation, and then the whole campus joined in. I finished some important development in about 85 years.
Q: What happened in the middle? A lot of optimization and improvement?
A: This is endless, right? New function, new function. In C language, a router is a class, and specific protocols such as IP are entities of a class, while NOS is multi-tasking. This is why Cisco is doing well, because you can add more things to NOS by simply adding another task.
Q: Speaking of Cisco, when did they appear?
A: In the spring of 1985, LenBosack (who was in charge of the computer facilities of the Computer Science Department and later became one of the founders of Cisco) and another person came to my office and asked me if I could access the source code of the router. I asked them what they could do, and they said they wanted to make some improvements and add more functions. I had other research tasks to do at that time, so I said yes and gave them the password. Cisco said it was founded in 1984, but I had never heard of it at that time.
Q: At that time, did you think the purpose of their code was to build a better campus network?
A: Yes. We have a meeting once a week and they are really doing something. When we decided to be a pure IP router, we cancelled the support for XNS, Chaotic Network and PUP. A year later, the router version they developed based on my code became the official router of Stanford. The job is good, that's all I care about. We are connected. It seems that at some point in 1986, I knew about Cisco, and we all knew about Cisco that Len worked for. They used their time at Stanford to develop code for Cisco. But this is not so bad, because similar things have happened in Stanford before. But Stanford decided that it was time to solve this problem. "You have developed the results at Stanford, and we want a piece of it."
Q: Who said that?
A: Just a general idea. Then I visited my lawyer and got the source code. Run the code in front of a lawyer. Let's talk about the operating system first, because this is the core of the router. They are all the same except the variable name. The lawyer said, "I am a lawyer. I think they are the same." . Let's take a look at the network data logblock again. It is divided into two parts. This is an improvement, but it's still the same thing. They changed something and added their own new routing protocol. If you know the internet, you can do it. I just did what I wanted to do, because it was the job given to me by the head and the demand of the department.
Finally, Stanford really decided that Len and his partners had to leave school and concentrate on Cisco.
Q: Apart from the patent fee of 654.38 million yuan, what else will Cisco give you?
A: One third of the royalties go to the school, one third to the department and one third to the inventor. I gave my third to the department, because in essence, all these came from a good research atmosphere. But Cisco never trusted me. There is a web page in Cisco that annoys me: "Sandy Lerner and Ryan Baussac are in love. They invented the router in order not to go out and chat on campus." It's really a joke.
I am the kind of person who is not welcome at Cisco. But these are all very interesting. I am passionate about everything I do, and I have also learned a lot from the operation of the company and how to become a big company.
Q: So you left Stanford for Sun after 20 years?
A: I left Stanford because it became more and more difficult to get funds from the school. I made extra money by consulting Sun. Mainly dealing with IMAP mail. Because they have a project called SPARCStation Voyager: using Solaris2.4, a notebook with fast matrix display takes up very little space. One feature of Voyager is that it can work in disconnected mode. You can disconnect the network, but he can continue to work. My job is to be an IMAP server and client, which can still work when the client is disconnected. This was difficult because IMAP2bis didn't support mail disconnection at that time, and I had to modify the protocol to support this and support low-speed bandwidth. When one of my colleagues left, his boss wanted me to join and continue the email part of this project. I think I'm 53, I went to college for a long time, and then I agreed.
Q: How do you compare business schools with business?
A: I often hit a wall in Sun's corporate politics, and I can never solve it well. When I was at Stanford, there was a rule: the best project won. Simple and direct. If your project is better than others', you will get the highest honor. But at Sun, things are different. Can produce the most correct software. There are a bunch of directors, deputy managers, managers, etc. I found myself involved in these struggles, and there were 10 people above me. I don't care about these, because I like to do the best projects.
So I brought IMAP technology, and by 1996, I had written IMAP servers everywhere in Sun. Then they decided that we should develop one. Then I invented SIMS(Sun Internet Mail Server), another type of server, and we ended up with thousands of inboxes on one server.
I have four patents about the Sims, and I have applied for 40 patents altogether. Everything else is related to peer-to-peer networks. Before coming to Sun, I did a lot of work as the CTO of JXTA project, which is an open source project of Sun.