The dynamic behavior of BGP is not well understood. Passive
monitoring of BGP update streams reveals a large number of
updates that are difficult to interpret. I'll present a
new formal model of the dynamic behavior of BGP updates
and claim that it helps to explain many of these observations.
The model highlights why we should expect BGP monitors at
different locations to collect very different update streams,
and why the concept of ``route flapping'' is so difficult to define.
This is very much work in progress.
Copyright. All Rights Reserved.