Marc Slemko wrote:

This will normally be a caching nameserver that may or may not be the same as the nameservers for the domain. In theory, it should be "close" to the client network wise for other reasons. Is it always? No. But it is a more reasonable metric.

Have any substantive empirical studies been perfromed that validate this assumption? The degree to which it is true seems to be the gating factor for the effectiveness of this method... Andrew Bender tns-inc.com