« Cows are magnets. | Main | Workshop: Flows and Networks in Complex Media »

August 30, 2008

The Internet is dead. Long live the Internet.

Having come of age during the rise of the Internet, it's hard to imagine what it was like to do science back in the communication dark ages without tools like email, electronic journals, Wikipedia, etc. Most of my research has some component that requires quickly communicating with people over vast distances [1]. This ease of interaction is all based on a few critical components of the Internet. First, the Internet is fast, in the sense that the Internet's routers and transmission lines allow information to get from A to B extremely quickly. Second, the Internet is navigable, meaning that A knows how to get to B using one of those quick routes. If either of these things failed, the Internet would quickly fall apart and people would go back to phoning and faxing each other [2]. Yuck.

What's not widely known is that there's a real danger that the second of these two won't hold in the future. If so, trying to quickly reach Google, your humble blogger or anyone else may become as difficult as trying to drive from New York to Santa Fe without the help of a map or road signs. The problem is that the system that makes the Internet navigable is fundamentally flawed, and it's not clear how to fix it now that everyone depends so heavily on there being a working Internet. That is, we can't just turn the whole Internet off while we move everyone over to the new improved system.

One flaw is that the system wasn't designed to have the whole, or even a large fraction of the, world online. It was always thought that we would roll out a new series of tubes [3] down the road, but then something happened: the Internet became wildly popular, and that idea was ruined. A second flaw is that the system assumes everyone is always honest. The Internet's navigability comes from, basically, a massive but highly accurate game of telephone. In the children's version of this game, errors are introduced accidentally, and everyone laughs at the end about how strange the messages become. In the Internet version, a malicious person can introduce an error strategically, allowing them to eavesdrop on other messages (a la the NSA) or hijack messages before they reach their destination. It was recently demonstrated that these kinds of attacks are, in fact, relatively easy to do. We've been lucky so far that these kinds of attacks haven't been more widely used.

These and other issues make it very clear that the future of the Internet (and my scientific productivity!) depends on designing a more robust system, to which we can smoothly transition while still using the current broken version. But how exactly would a better system work? Earlier this summer, I coörganized a mini-workshop at SFI with some folks from CAIDA (Cooperative Association for Internet Data Analysis) based at UC San Diego about exactly this question. The attendees were primarily folks from the Internet-branch of the network science community, and the talks were focused heavily on alternative ways to make the Internet navigable. The result of the meeting was not a solution to the problem, but rather a set of questions that we think probably need to be answered before a real solution can be made.


[1] Usually, this is because my collaborators are remote; one of them I wrote two papers with before meeting him in person for the first time earlier this year.

[2] Increasingly phone calls depend on the same technology (packet switching) that runs the Internet, but originally, phone calls depended on a different kind of system (circuit switching), which guaranteed the delivery of information (i.e., no garbled conversations because of network congestion) but was significantly less flexible.

[3] This was a naive view, of course, but it's hard to make accurate predictions, especially about the future. The future was supposed to be based on something called IPv6, but if everyone used IPv6 today, it would break the Internet even faster. Fortunately, or unfortunately, almost no one uses IPv6 and it seems that no one is really planning to, either.

posted August 30, 2008 08:24 AM in Networks | permalink