Glitches happen, and this one illustrates how rankings with big differences in spam volume are robust anyway.
A format change in an ancillary data source detected through consistency checks caused recomputations in selected rankings for September, October, and November 2013 in
Classic.
(Cloud. was unaffected).
The old versions are preserved as v1 rankings, and the differences are visible for these overall rankings:
| Geography | Sep 2013 | Oct 2013 | Nov 2013 | |||
|---|---|---|---|---|---|---|
| CBL | PSBL | CBL | PSBL | CBL | PSBL | |
 World |
CBL** | PSBL | PSBL* | PSBL** | ||
 BE |
CBL | PSBL | PSBL | PSBL | ||
 CA |
CBL | PSBL | PSBL | PSBL | ||
 TR |
CBL | PSBL | CBL | PSBL | PSBL | |
 US |
CBL* | PSBL | PSBL | PSBL | ||
 Countries |
CBL | PSBL | CBL | PSBL | CBL | PSBL |
| Medical | ||||||
| World |
CBL** | PSBL | ||||
| US |
PSBL | |||||
| Countries |
CBL* | PSBL* | ||||
| * Completely unchanged in rank order | ||||||
| ** Unchanged except for dropout final rank | ||||||
So the most noticeable rankings, for World, were almost completely unchanged in the top 10. And the U.S. CBL rankings were completely unchanged.
The affected 32 rankings had changes in volume due to this recomputation, out of 54 total rankings for those 3 months. For rankings that changed, the ranks that changed were mostly for entries with very similar volumes. The other 22 rankings were completely unchanged, and are shown as blanks in the above table.
The specific cause of this recomputation was a change in the format of the daily Team Cymru data we use to map netblocks to ASNs.
-jsq
