The process was reasonably random - the survey was distributed to the global Java User Groups & I'd say over 2000 responses is something worth paying attention to statistically (yeah, I know - out of x Million Java developers it's still a tiny sample set).