analysis of BGP tables


data is clean, easy to get, parse
appears a reasonable substitute for traceroute data
fluctuations
NLANR's archives corrupted Dec.2000-Feb.2001
random daily fluctuations (~3%)
if a single /8 is not announced - 0.4% IP addresses 
misconfigs
routing loops (2 loops in 91 pref-peer pairs)
discrepancies
e.g. multi-origin prefixes, 1.3%
caveats: extremely sensitive to peer choice