On Power-Law Relationships of the Internet Topology

The basic methodology and contribution of this paper is to examine aggregates from routes collected by BGP over a period of approximately one year. Given this data set, the authors argue that the relevant properties of the internet graph are best described by power-law distributions rather then simple averages. For instance, in this work, they describe three power-law distributions: between router out-degree and router rank, router out-degree and frequency of that degree, and the size of hop-neighborhoods. The authors that their results show a sort of invariance in the basic structure of the internet even though the network changed significantly during the time during which the measurements took place.

This paper seems easy to criticize, both typographically and methodologically. I do want to try and appreciate what the authors are attempting to accomplish; this seems like a very difficult type of paper to write, since no matter what your conclusions and data you will probably get attacked. All that notwithstanding, my main question would be to consider how well this work has aged; the internet has certainly grown by much more then 40% between 1997 and the present, and the authors' conclusions would be considerably more convincing if they held between then and now. It appears from a cursory googling that the authors did publish a sort of follow-up in 2003 where they claim that these laws continued to hold over five years.

While I would be a forgiving reviewer towards the actual data they present, I do take issue with some of the final sections, where the authors wax poetic about "finding order in chaos." They also present some completely uninterpreted results about the power law of eigenvalues. While they cite some work which I'm sure has very interesting interpretations of why we should care about the eigenvalues, this section seems particularly unmotivated.