Amazon S3 is a web storage service by Amazon. Essentially, you get buckets you can store objects (files) in, and others can access those objects directly through the S3 service or via the web. This is particularly useful for hosting images and video...the storage and bandwidth go to their server farms, easing the load on your own server. Space and bandwidth are rather cheap ($0.15/GB for storage and $0.10/GB for bandwidth as of this writing), and you pay only for what you use.

Anyway, one little trick which I just finished setting up (which is what prompted this post) is that you can set up DNS CNAME records to refer to Amazon S3, so you can have files hosted on S3 which appear to be hosted on your own server.
Like this:
http://media.arklyffe.com/emaganim.mp4

I'm also working on a C++ library for interfacing with S3, and a set of command line tools for working with S3 buckets and objects. It's not quite in releasable condition yet, but I should have something up soon.