Freek Dijkstra was kind enough to provide a graph of the relations between networks
at the end of the experiment. I have reproduced it in figure . on page . Using
this graph, it can be seen that not every network should be reachable from all other
networks, assuming correct policies in place. For instance, Rudy does not buy transit
from anyone, thus his routes should not be propagated beyond his direct peers. How-
ever, in practice Dick did not have filtering in place and announced Rudy’s prefixes
to micheal and me after which we would announce them to our customers. Another
example is Niels. While he does peer with me, he does not buy transit from me or
anyone else. This means he would be reachable from my network and the networks
of those who buy transit from me, but not from those who peer with me without
also peering with Niels. Again, because not all routers were configured with correct
policies, his routes were propagated beyond where they should have.
Note that I also do not buy transit from anyone. This was a deliberate decision: by
only peering and providing transit, my network would be a so-called ‘ Tier-’ network.
In this graph, given correct routing policies, Rudy, Dick and Niels would be left with
a partially unreachable network. They would need to buy transit from someone else
to correct this situation, leaving Michael and me as Tier- network providers.