I agree - while this new feature from S3 makes sense for a segment of users, just like in normal S3 usage, the complexity is left for the client application to

I agree - while this new feature from S3 makes sense for a segment of users, just like in normal S3 usage, the complexity is left for the client application to deal with. This is a non starter for most companies. Full Disclosure - I am an employee of Nasuni. We just (today!) launched a versioned, encrypted filesystem sitting on top of cloud storage (including S3) (http://www.nasuni.com http://www.nasuni.com). You are right to point out that this implementation put the burden on the application developers. The fact that each version is a complete copy of the object means that your storage growth could be massive depending on your write pattern. For example, take MS Word - the number of changes generated when editing a file means that if you were to use this feature directly you could end up with hundreds of versions of that single object. This is a good feature that will apply to certain types of S3 users, but those S3 users will not likely fall in the class of generic file/file system application users. Amazon is the only cloud vendor that supports this right now - if you plan on using multiple cloud storage vendors you'd need to consider a design a cloud-neutral application yourself to handle this.