Summary: Tussle in Cyberspace

This paper basically argues that we need to take into account the economic
realities of the Internet when designing new solutions. When the Internet
began, all of the participants had the same goals: hook up computers located
at research institutions to each other and see what sorts of neat stuff you
could build. Today, the Internet's participants do not all share the same
goals. Therefore, it is impossible to build solutions that will satisfy
everyone all the time. The best we can do is build flexible solutions that
remain workable even when being used by people with different interests in
mind.

The main principle expoused in their work is that designs must allow for
varying outcomes, depending on the goals of the participants. The tension
or "tussles" that exist between players with different interests should not
stretch the design. Instead, people should have some way of expressing their
goals within the confines of the system such that they can influence its
behaviour without violating the design. The authors also explain that
systems should be designed so that tussles over one aspect are as isolated as
possible.

The arguments presented in the paper are a little abstract. Fortunately, they
present a number of real world examples. They point out that the Internet
should have been designed to make IP addresses very portable from one network
to another so that people can easily switch ISPs. Since customers wouldn't
be locked in with one ISP, the competition between ISPs would be increased,
and the overall level of service improved. Unfortunately, they don't explain
how exactly this would work while keeping routing tables small, or why it is
even necessary given DHCP, etc.

One could argue that the conclusion presented in the paper is not useful when
taken to the extreme. For example, consider DoS attacks, which allow a small
group of users to take down a site at the expense of seriously perturbing the
network connection of innocent bystanders. The logical extension of the
paper's argument is that the network should provide some way for a small
group of users to vote a site off the Internet. This way, the tussle between
the site and the attackers would happen within the confines of the system and
wouldn't affect innocent bystanders. However, it's clear that such
a "feature" would be completely undesirable.

The general problem is that some people's goals are inherently illegitimate.
Network designers should not be obliged to build mechanisms to support such
goals. However, without doing so, there will always be users that will
devise attacks on the networking infrastructure to attempt to carry out their
ends.

I though this was a fun and interesting paper. It explored a design
philosophy more than a particular idea. In many ways, it is like the "end to
end" paper with which we started the course. However, the ideas presented
in "end to end" seemed more useful and concrete than those discussed here.
Received on Sun Nov 26 2006 - 22:50:33 EST