What stops speed tests being done that use real sites for their tests? I could envision using a random YouTube/Netflix/Hulu video for a test to be done across

What stops speed tests being done that use real sites for their tests? I could envision using a random YouTube/Netflix/Hulu video for a test to be done across several providers at the same time. Add some legal torrent and FTP servers as you're at it. That way you get results that can be compared directly. For end user results, you really don't need to be that thorough when comparing results (after all, if you're doing at different times on different you are already adding unhandled variables) you just use any random video from one of those providers.