I think that the transmission technologies were the things which laid most of the groundwork in the ‘90s, so figuring out how to consume usefully the computer cycles and communications bandwidth that we know how to produce now is going to be the big challenge. That’s again a question of

reducing the crisis of complexity.

The evolution of the Internet to date, when we talked about the milestones, was shaped by the changing population going from network researchers to university researchers to industrial to the general population. So I think that the future is going to be shaped by those demographic changes and universal use, particularly by people that do not use European-style alphabets.

 

The other thing is that machines which aren’t things that you type on or even supervise are going to be connected, and they will chit chat. We used to think that having an Internet address on your car or your refrigerator was silly but I think that probably in about ten years from now, we’ll expect to see multiples of them. I see no reason why, for example, if I take my car in to be fixed, it has a jack where mechanics can connect up and they can find out what’s wrong, but I can’t. Why doesn’t my car just tell me directly, rather than telling the people that want to charge me a lot of money?

 

In the past as you added more transistors to a chip, the first thing that happened that every chip had a computer on it, and you had computers in microwave ovens and so on. The next thing that happens when you add transistors is that you get the ability to communicate over a network interface, and I think wireless is important in giving us a way to connect almost anything to a network. So I would expect that a car that I would buy five years from now would be able to give me status reports about what it thinks need to be fixed, whether the brake pads are wearing out or whatever, and it won’t need the jack any more – it will use wireless. So the “Internet of things” I think is the big thing.

 

The real challenge that we’ve come back to is figuring out how to integrate the control systems for all of these network devices that will talk to each other. Because in order to make the Internet of the future work, we’re going to have to make it less complex to have a whole bunch of things talk together.

Right now, complexity feels like the thing that will limit us. We tend to add computers and networks and so forth up to the point where they get to be too much of a headache to add the next feature. I don’t know about you, but there are lots of times when I haven’t tried new software or new features just because it was too hard to learn them. So reducing that complexity is the way to keep the expansion of the Internet alive.

I have three children, one’s six and I have twins that are four years old. My six-year-old - when he learned to read – would see buses going by with URLs on the side and he would say ‘Dot com, Daddy did dot com’.

I actually didn’t do .com per se. I did the DNS technology and left the choices of the domain names and the administrative stuff to people like Jon, so it’s not really true that I did .com. I think it’s a funny kind of thing and now my four-year-olds are starting to do the same thing.

One of the things I think is most amusing, again another .com story - I designed the technology to allow you to do domain names but I didn’t define which domain names we would use. For example the people in the

UK prefer .ac (academic community) as opposed to .edu. I had made none of those choices, because I didn’t want to get involved in those politics. Then when it was time to define them, there were all sorts of bad ideas about how to select domain names. So that, for example, there were was one famous individual who said ‘In all domain names, the most important part should be a country code’, instead of having .com and so forth. I’ve always been in the mode of thinking that experimenting when you can afford it is a much better idea. So I said - well let’s try both and see what works. So the only reason we have .com today is that finally people got tired of arguing with me about that and said ‘Well we’ll try it, but no-one will ever use it’. If they had known what was going to happen they would never have agreed!

It wasn’t that I thought the .com was the right idea, I thought that trying both country codes and generic and see what worked was the right thing. I tend to think that that kind of experimentation was a good idea, and I wish we had done more of it!

Another thing was somebody said that you should have your network name instead of a country code to be the most significant part of a domain name. Because you have to know whether you’re part of the ARPANET or not in order to be able to talk to the NSF net or MCI - and we managed to not do that. So I think there are several disasters that we have managed to avoid.

• Well, for example there’s Bob Kahn who is well known as being a co-inventor of TCP. But in reality, a lot of the work that he’s done that people don’t know about has probably been as important. For example, he contributed to what we call the Gigabit Program in the

US for doing high-speed nets at the time. It’s kind of funny, in the 1970s I was working on a high-speed network technology that was a megabit. Then, in the early ‘90s a gigabit was high-speed technology - and today you have to be multiple gigabits to rate. Bob also works on the management of intellectual property and how to deal with similar issues. He’s been a consistent influence over the years so he’s one of my favorite contributors to the technology.

· There are the usual famous people like, Vint Cerf.

· Doug Comer did a lot early on, just to write a book about the Internet technology, and make it more accessible to people. He did a bunch of work with CSNet and so forth. In the early days in particular, the important thing to do was actually figure out how you could make technology that you could spread. Nowadays people talk about who owns it and how to regulate it – for a while it was just important to get it invented and accessible.

· Van Jacobson who made a bunch of technical contributions.

· Carver Mead taught a class at CalTech that I managed to attend via Dave Farber’s influence. In those days, designing chips was incredibly hard. You had to use a batch computer to do circuit simulation, colored pencils to draw layouts, etc. While Carver is most famous for all his integrated circuit work and general brilliance, he also was a key in getting the MOSIS project going which let people build chips collaboratively over the net.

· Dave Clark at MIT made also a bunch of interesting technical and also management contributions.

· The late Jon Postel - Jon’s contributions were, in my view, more of the management, organization and collaboration side of things, rather than in any particular technical thing. Getting people to use the technologies and to work together on it is often as important as having some brilliant idea.

· I guess last, but by no means least, Marshall Rose has had more spectacular successes and failures than anyone else I can think of. There’s also his sense of humor, irony, and sarcasm.

· Anyway, there’s an assortment of people.

Well actually I know him from when I was at UC Irvine, or at ISI.

Marshall was an undergrad. When I was a Grad. Student and we worked together somewhat clandestinely to get UCI connected to the Internet because it turns out that the people at UC Irvine had people studying what was called at the time the Social Impacts of Computing and they wanted to be able to do a controlled experiment to kind of watch as you connected the university to the Internet to see what was going on. You didn’t want to introduce it without first doing sort of an ecological study to see whether or not it was a good idea and so forth. And so we just sort of did it one night.

I have to backtrack a little bit. When I was in MIT, I was working with Nicholas Negroponte and the people at the Architecture Machine, now the Media Lab. Because we were a small, under funded project at the time, we used a lot of small mini-computers that people would donate, rather than the big machines and timesharing systems of the time. So from the beginning, I’ve always been interested in working on the organization of multiple machines - what you would call, in general, distributed computing.

When I was at UC Irvine, I worked with Dave Farber on the Distributed Computing System or DCS. That contributed a bunch of the key ideas to Local Area Networks, in particular the token ring hardware and operating systems. We also needed a distributed naming system.

At ISI, as there’s the DNS.

One of the other things I did at ISI was work with people on things other than the DNS. At the time when people changed from the NCP protocol to the Internet there was a chance to re-think all the protocols that were in use. In the case of mail, the ARPANET combined mail and FTP inside the FTP protocol. For the Internet, we defined a separate mail protocol, originally called the Mail Transfer Protocol (MTP), which luckily got simplified to SMTP. I remember late hours trying to get the first email server written to test the new mail protocols, and even more “fun” when the ARPANET protocols were turned off, and I was suddenly doing production, not protocol research.

Anyway, I think I had the first SMTP server on the Internet.

After working at ISI in the university environment till ’95, with three years at ARPA, I worked at several startups, one of which was @Home, the first company that did Internet over cable on a large scale – some say @Home’s marketers invented the term “broadband”. And though the company has gone out of business, Internet over cable has never been better. In the

US, cable’s got more of the broadband population than DSL. In other parts of the world DSL is more important than cable. To some extent I don’t really care. The important thing is getting broadband out there because I think that’s going to be creating some of the next generation of uses of the technology. The people that I interact with, for example, don’t necessarily think that it’s critical to exchange MP3 files, but some of the users will. So I think Internet broadband, DNS, the mail protocols and LAN I think were the high points.

I think there are two ways to look at that, - one of which is technology and the other is more social - so I kind of do both.

· Obviously the 1983 adoption of the Internet technology was one of the key milestones.

· With regard to my work, the invention of the DNS protocol, 1983 was also a milestone. It’s now the 20th anniversary.

· 1986 was the year when we saw the first hosts that actually depended upon the DNS - they didn’t have a host table any more. That led to the idea that a lot of the configuration information could be embedded in the directory system in the network and that it wasn’t an optional feature any more. It was part of everyday life.

· Another interesting milestone in that vein was the use of MX records to route mail. It was really the first application where we were using this distributed database for stuff that wasn’t just host to address mapping, and it was the first addition to the DNS by folks other than me. It was something that was fundamentally different. Over the years we have seen over forty other types of data added to the DNS.

· I think around the early ‘90s, when people did DHCP and some of the automatic configuration stuff, all of a sudden the network became more dynamic, rather than a static configuration.

· And then, of course, around 1995 the Web was the killer application that everybody in the early ‘90s was searching for.

So those are important things from the technology aspect.

But from the social aspect, the population of the network has always been changing. Originally it was for network researchers in the ‘60s. The ‘70s was the decade of research and education - the university community - not necessarily network researchers, not necessarily computer scientists - starting to use it. And then came the commercial users, the companies. And then the non-US general English/Latin-alphabet speaking population of the world.

Now you’re starting to see big growth in two other populations: non-Latin character sets and languages (Chinese, Japanese, Korean, …), and also an “Internet of things” - PDAs and instruments and so forth - machines that people will never directly interact with, but which will provide information.

These –milestones - the change in the population and the changes in the usage - are equally important as the technology milestones, because I think the one drives the other. 

DHCP: The Dynamic Host Configuration Protocol. Is an Internet protocol for automating the configuration of computers that use TCP/IP. DHCP can be used to automatically assign IP addresses, to deliver TCP/IP stack configuration parameters such as the subnet mask and default router, and to provide other configuration information such as the addresses for printer, time and news servers.

The prior paragraph is pretty incoherent, so if you want I’ll rewrite it. The basic idea is that 1986 is the year where we saw hosts that depended on the DNS beginning to appear.

The most important characteristic of the Internet is that users and designers are free to combine the protocols and features in a variety of different ways.

It’s nice, for example, that TCP can be used to distribute routing data or it can be used for the web or it can be used for email or it can be used for Telnet. It is nice that the DNS can be used as distributed database for new applications.

Providing features that support the immediate demands, but also can be combined differently in the future, is key. Every piece we implement should satisfy some need (to see if it works) but also bring (possibly unknown!) future capabilities one step colder.

We haven’t always succeeded in building things that way - sometimes we build things that are designed for a single purpose and that aren’t very easily used for other purposes, and I think that’s sort of short-sighted. For example, standards organizations feel a lot of pressure to show results and so sometimes they artificially try to narrow the focus of what they’re doing and make things go faster. I think that’s kind of a mistake when what you do is to just narrow the focus to exactly what you need to do today without thinking about how you provide a capability that can also be used in other ways. Oftentimes, just focusing on the specific problem of today is plain short-sighted and leads to bad designs that just make things harder in the future. Standards organizations also tell people not to use a feature and instead wait for a new protocol which will be available “real soon now”. I’m much more in favor of giving everyone a set of tools, and letting them try things out, and then judging the results.

I mean it’s nice that the Internet can be used to implement VPN but it would be a mistake to force everybody to use a VPN.

Another key characteristic is the fact that, at least potentially, there is no restriction on connectivity on the Internet.

After graduating from MIT, and working a bit, I went to the

University of

California at

Irvine for my PhD. Dave Farber was there at the time as a Professor. That’s how I got remote access to the ARPANET, just using a terminal on a dial-up to assess a TIP. We also built both the hardware and software for a local area network, and the ARPANET connection helped us see what others were doing.

After a while, one of the big ARPA research places, called Information Sciences Institute in Los Angeles, had money to hire grad students and so they hired me, even though ISI was part of the University of Southern California and I was going to a different university. They kind of overlooked that difference but I think they’ve closed that loophole now, so you won’t find anybody else who has the same story. Now you have to be a USC student to work at ISI.

Anyway, that’s how I really got involved in doing all the Internet stuff, because at that time it was late in the ARPANET, and people were building the next generation, or Internet. ISI is where I did the DNS and various other Internet things that people know me for.

I think I started working with the ISI people in 1978 and I became a formal employee of ISI in 1982 when I graduated and I got my PhD.

I went to high school in

Boston, and MIT had various free summer classes for high school students that were taught by volunteers, often MIT students. I went to two, both taught by MIT students who are now professors, what a bargain! One of them was about computer programming. I learned a bit about programming and that’s where I first had contact with a computer, an IBM 1620.

The machine that we were using was part of the Department of Aeronautics and Astronautics - aero planes and stuff like that. The secretary there asked me one day whether I would like a key to the computer room and I said ‘Sure’. She said, ‘Are you a student?’, and I said ‘Yes’. I didn’t realize until later that she meant “was I -an MIT student?” as opposed to “was I a student?” So they let me have all the access I wanted to a computer, mostly late at night, since the daytime had serious MIT work going on. That would have had to have been ’65 or ’66.

When I graduated from high school, I went to MIT and continued to use the same computer, now as an MIT student. The Aero people were used to having me around by then and I even had an MIT ID. I remember one funny incident where I was using their computer to do a simulation of a nuclear reactor (basically what’s known as ray-tracing today) because I thought doing the math was too hard. A couple aero students and a professor started talking to me about it, since they found it interesting, but told me that it was impossible to “fly” a reactor that large. They also asked me to help them with airplane simulations, but were amazed that I knew nothing about airflow and the related equations