At a first pass, I'd try to implement such a thing with a Distributed Hash Table-like situation. Generate chunks in some fashion (either one chunk per file, or

At a first pass, I'd try to implement such a thing with a Distributed Hash Table-like situation. Generate chunks in some fashion (either one chunk per file, or glob files together up to N megabytes), take a hash of the chunk, and then let people fetch it. All the Internet Archive needs to do is then keep redundant copies of the hashes.