Before I decide to work on it sometime in the future, anyone else interested in creating a LocalFileRepo for Amazon's API? Unless someone corrects me, the best method of dealing with Amazon S3 for storing images would be to make use of S3's API, rather than mounting buckets onto the filesystem. The former should be more reliable (^_^ trying to use a mountpoint will probably drive someone up the wall like NFS does for brion), and also using the API should be more reliable for handling multiple buckets, since as I recall the Amazon docs say that buckets can only hold up to 5Gb each.

Though considering the large things to deal with for multiple buckets, and the fact that the best methods will probably also have some url redirect handling as well to keep the standard urls, it might be best as an extension rather than put into core.