On Thu, 31 Mar 2005, Florian Weimer wrote: > Yes, the selection of criteria could be biased. Or Telcordia compared > apples and oranges when it compared Verisign's 100 ms to DENIC's > 200 ms (or what the actual numbers where). Yeah, I was a little curious about the composition of the latency number as well... A heavily-splayed anycast deployment should have influenced that number favorably, I'd have thought, but apparently not. It's my assumption that they ran pings (of some unknown duration) from some unknown number of locations, to each of the currently-operated server addresses, and combined (averaged?) the results somehow. But I'd certainly be curious as to their actual methodology. -Bill