goodgirl discusses it
here, or you can just go straight to
Google.
2.3 Automatic Filtering - <filter>
The <filter> parameter causes Google to filter out some of the results for a given search. This is done to enhance the user experience on Google.com, but for your application, you may prefer to turn filtering off in order to get the full set of search results.
When enabled, filtering takes the following actions:
* Near-Duplicate Content Filter = If multiple search results contain identical titles and snippets, then only one of the documents is returned.
* Host Crowding = If multiple results come from the same Web host, then only the first two are returned.
How do they determine that they are from the same host? Well, the Class C and nameserver are the easiest, which is why we provide multiples of both on all our packages.