February 28, 2004

The Free Market Problem of Network Abstraction

It is extremely common to use abstraction as a tool to simplify complex computer networks. For networks of any appreciable size, it is highly common to simply draw a cloud shape for the majority of parts not under examination, slap a label on it and move on talking about the local bits that you actually care about. The biggest of example of 'cloud technology' is the Internet.

Abstraction is useful in that it permits us to talk about things too complex to hold in our heads if described in its actual detail but the risk of abstraction is that the stuff that is left out might just be important. For the politico-economic analyst talking about the Internet, 'cloud technology' is a nasty enemy. It leads people to a great amount of fundamental error. It is perfectly proper to say that no one person owns the Internet. It is absolutely false to say that nobody owns the Internet, as false as saying nobody owns IBM or Microsoft. Yet people do it all the time.

The Internet is a set of agreements between the various private property holders to, broadly, carry signals between each other both as endpoints of traffic and as pass-throughs, intermediaries. It is all contractual, all perfectly capitalist/free market friendly, yet because nobody can hold the sum total of these agreements and they are negotiated and consummated automatically, some act as if they do not exist and speak of the Internet without taking into account its private, consensual nature.

ICANN some say, runs the Internet. They do not. What has happened is that a critical mass of the property holders who actually own pieces of the Internet have agreed for ICANN to create rule templates to speed the progress of negotiating new contracts between the various owners of the networks.

This has broad implications for net governance and for appropriate reaction to misgovernance. I believe that it is inevitable that there will be a point where somebody who is writing these contracts and embedding them in router firmware will fundamentally err and attempt to push the owners to do something that a viable portion of the Internet infrastructure simply does not accept. When that day comes, the Internet will fragment. In software development terms, this sort of fragmentation is called a fork (think fork in the road). When the Internet forks, it can fork in a way that preserves interoperability, simply changing contracts in the rebel portions and ensuring that bits can pass between the two new networks or it can fork ugly and traffic interconnection will either become slow and congested or fail outright.

The truly funny thing is that from a political standpoint, if there is a comprehensive engineering solution to fork with minimal cost, the incentive is to never push things so far that a fork will ever come about. The 'internet leadership' will mostly consist of determining where the collective owners of the Internet, the shareholders if you will, want to go and to rush ahead of the parade and pretend that they are choosing the direction.

Posted by TMLutas at February 28, 2004 09:16 PM