Not that I doubt that, but how would you even go about measuring such a thing? Obviously, there's a cost to using google version too (DNS lookup + one more poin

Not that I doubt that, but how would you even go about measuring such a thing? Obviously, there's a cost to using google version too (DNS lookup + one more point of failure). It would be nice to have real numbers for what percentage of visitors actually have the libraries cached. Remember, that the files are versioned and they're also serving less popular libraries than jquery. My guess is that it would be slower to use the google version for some library/version combos. The question is which ones.