Heather:
This prior question from you (November 2008) was recently brought to our attention.
Sorry about this delayed response, but we thought it would still be worthwhile to share
pointers to some work that we are doing at NIST which relates closely to your question.
Earlier Bill Woodcock provided you with a link where the actual discrepancies
are listed. Our work at NIST focuses on the statistics of such anomalies,
with the intention of: (A) generating score cards for accuracy/consistency
of various registries, and (B) to glean the "good" data from what is available so that
BGP robustness algorithms that rely on the data can work more effectively.
We have done an analysis of registry information (RIRs, IRRs, RADB) and compared
it with that from trace data (RIBs, update history) from RIPE-RIS and routeviews.
We generate a variety of statistics on a per RIR basis (ARIN, RIPE, etc.) regarding
whether announced {prefix, origin AS} pairs in updates correspond with those in the registries.
We also report on whether the registered objects (NetHandle and AShandle in SWIP format
and inetnum, aut-num, and route in RPSL format) appear self-consistent or not.
We also looked at the NetHandles in ARIN that contain origin AS information, and have
performed comparisons of those with what was historically seen in BGP updates for
prefixes belonging to the ARIN region.
A variety of results and discussion related to all this are presented in this set of slides:
http://www.antd.nist.gov/bgp_security/publications/ARIN_NetHandle_OriginAS_Analysis.pdf
You may also look into a presentation we made in January at NANOG-45.
There the focus was on BGP robustness algorithms that make combined use of
filtered "good" data from registries as well as long-term trace data.
http://www.nanog.org/meetings/nanog45/abstracts.php?pt=MTE5NSZuYW5vZzQ1&nm=nanog45
Here is a link for a detailed published paper related our NANOG-45 presentation:
http://www.antd.nist.gov/pubs/NIST_BGP_Robustness.pdf
(This paper was published in the Proceedings of DHS S&T CATCH 2009 conference.)
Please let me know if you have any specific questions concerning the above.
We are very interested in receiving feedback on how this work can be made
more useful from the perspective of what ISP needs are.
Sriram
K. Sriram +1 301 975 3973
http://www.antd.nist.gov/~ksriram/