A minor detail, but I am surprised about the whole "it was hard to figure out when someone moved a file" thing. I always assumed the actual file on disk was abstracted away into a database record, that specified the folder etc. Were they actually using S3 Bucket/Folders for file structure on Dropbox itself?
It was more of a lack of abstraction than anything. The server reflected more or less what the OS presented -- an add and a delete rather than a move. Putting the pieces together later to present better information to the user was difficult. Not sure what they do today, but I remember many talks about ways to put that abstraction in place as close to the OS as possible. Reconstructing multiple, quick, renames/moves after the fact proved to be somewhat difficult from the frontend, which Aston worked on.
