Hi Eric, although I agree there are biases in these tests, it would be practically impossible to run any such test that wasn't biased due to the sheer number of variables.
Further, such tests make little sense in the first place. In comparing the end-user equipment, what's the point? We already know the max capabilities of these pieces of equipment. Any performance short of that just measures limitations further up the road, which are not deterministic.
Even if you could conclusively say that, for example, a particular Cox@Home-Omaha connection is faster across all time and space compared to SBC DSL in Fort Worth, what good does that information do anybody? You will never have those 2 options to choose between so what value does the comparison have? Also, it is entirely possible that a different connection on the Cox@Home-Omaha network would produce different results.
As a side note, they may have chosen these particular downloads because they are all tagged as non-cacheable. Obviously from the results, caching played no role in the outcome.
dh |
