This paper studies the large-scale behavior of routing in the Internet. Instead of focusing on information propagation
inside the network, the author attempts to observe the routing dynamics from an end user's point of view.
The three key main ideas of this paper are:
1. The likelihood of encountering a major routing pathology more than doubled between the end of 1994 and the end of 1995,
rising from 1.5% to 3.4%.
2. Internet paths are heavily dominated by a single prevalent route, but that the time periods over which routes persist
show wide variation, ranging from seconds up to days.
3. Internet became more asymmetric at the end of 1995 than 1994.
In summery, during 1995, Internet routing became less predictable in major ways.
I rate the paper's significance as significant contribution(4), because very few research work were devoted to quantitative
study of the Internet's large-scale behavior prior to this paper. Understanding the real behavior can help us make correct
design assumptions, which are very important for the success of future improvement on Internet architecture.
This paper provides a lot of experiment data, which, together with sound analysis, make the paper looks quite convincing.
However, the credibility relies on the assumption that the selected sites are representative for the entire Internet. To
show the assumption is valid, the author argues that the selected sites include a non-negligible fraction of the AS's which
comprise the Internet, and are weighed according to their real occuring possibility. The argument was sound at 1994, but the
situation has been changed since then: most current sites are commercial websites. So the original list, which are most made
up of universities, is not representative now. To repeat the same experiment correctly, we have to select new samples.
The limitation of the paper's approach, "end-to-end measurement", is that it has the difficulty of compounding effects at
different hops at the network into a single net effect.
From this paper, Internet researchers and builders can have a clearer idea of the routing behavior of the Internet in 1995,
which include pathological conditions, routing stabilities, and routing symmetry.