Notes from the 10/2 discussion on the "Delayed Internet Routing Convergence"
paper. 

Andreas Terzis

-The experimental data is good but the theoretical model is somewhat
simplistic
-The real network is not a full mesh and the existence of policies might
actually reduce the convergence time by reducing the number of feasible
paths
-However Tier-1 ISPs are connected in a full mesh and therefore the results
might apply in that subgraph. 
-Worrying about length of convergence time should not be the only 
concern. We should also care about the ability to forward packets during
this time. In that respect look at [1]
-[2] is one of the proposals to reduce the convergence time. 
-It would be interesting to know what are the routing mechanisms used by the
telephone network.