A movie goes about two hours. If you enjoy the movie, you will not notice that you just spend two hours. It will touch you, and you will have a lot to think about – at least for a good movie. A talk of Vinton G. Cerf feels exactly like a good movie.
March 10th, 2006, University Saarbrücken: In a 300-people auditorium, people are standing in the back, and google employees give away google notebooks (er, unfortunately only the paper ones)
Vinton Cerf is currently vice president of google and “chief internet evangelist”. He is often seen as one of the fathers of the internet. Nobody doubts that the winner of the 2004 turing award knows what he’s talking about – in the first few seconds, he already won the audience by starting with “Vielen Dank meine Damen und Herren…” – he once lived in Saarbrücken.
Since we arrived just in time, I didn’t get a regular seat but I am sitting in the back of the auditorium on a small table. I only see one third of the transparencies, but since he talks, this doesn’t matter.
During his talk (“Tracking the Internet into the 21st century”), he lists the challenges that we currently face with the structure of the internet on a technical level. Dr. Cerf is really enthusiastic about mobile devices – the net currently consists of 360 million hosts, but there are 2 billion mobiles around. Mobile applications have a huge potential.
The Internet Protocol (IP) suite has laid the foundation of the internet success story. I never thought about why in detail, but Dr. Cerf points me to the right direction: the IP suite is a layered protocol stack. This enables applications to use e.g. a TCP connection to transport data (like your browser uses a HTTP-connection to the webserver). However, only the webserver and your browser needs to understand the HTTP protocol. The network devices in between just transport the information from the sender to the recipient. This sounds really simple, and it is. But this design enabled the development of applications that reuse existing network infrastructures – which is why the internet has grown to its current size. (Note to self: Remember that end-to-end protocols and layers are good for open, scalable systems.)
Dr. Cerf continues to compare the design decisions that were made in the seventies with today’s requirements. Often, the protocols are not capable of handling today’s requirements, but it is hard to change the environment. If you take IPv6 as an example, the protocol suite offers a lot of advantages. But you need to change the way you set your infrastructure up, and therefore organizations are reluctant to implement it. And home users are just not interested in the deployment off something that has no direct benefit to them.
You can find more hints in my raw notes. He continued to talk about mobility, security, threats, routing, the DNS system, intellectual property protection (“hey, content industry, embrace the new technologies!”) and the lacking competition in the broadband market.
If Dr. Cerf would be able to start over, he would do another design around the end-to-end connectivity principle. He would add authentication and flexible VPN memberships, incorporate IP mobility, confidentiality and think about flexible binding of DNS names to mobile IPs. But currently, he is involved in the development of the InterPlaNetary Internet (IPN), an attempt to reuse existing satellites to communicate with other spacecrafts. Basically, you cannot use TCP/IP because e.g. congestion control will not work for roundtrip times of 40 minutes. The project needs to deal with communication endpoints that become unreachable for hours (e.g. because a satellite is hidden behind the moon). This sounds really interesting – maybe it can be used for terrestrial applications as well.
After one and a half our, the audience asked questions for another half an hour, and google funded a nice reception. If you have the chance to see Vinton G. Cerf, go there and enjoy the performance.