When I tried to download Selenium standalone jar file. I realized that they it is not a simple Apache file system. For example, the URL looks like:
And if you go xxx.com/2.44, there is no content, but you go one level up, it will return you an interesting XML file. Obviously the part after the com/ is the key in the XML file and it is hosted by Amazon S3.
The reason that Google uses S3 instead of old-school Apache directory is because they have more options around security and permission. Probably down the road, I need to learn how to use S3 to share file with the public with customized permissions.
Here is more information about the S3 Rest API.