It is possible for files that have been staged to be overwritten by a new upload, if that new upload uses the same "unique" ID as the previous upload. If this happens it is also possible that the sweeper that moves these files into place will move a partially written form of the file into the archive; a Bad Thing (tm).
Currently we depend on Resumable to generate the "unique" identifier for the upload. It seems a better solution to have the client prepare an upload, and have the WS generate a (truly) unique ID in response. This way we can guarantee no collisions between uploads and prevent a second upload with the same ID from starting if the file has already been successfully staged.
It is possible for files that have been staged to be overwritten by a new upload, if that new upload uses the same "unique" ID as the previous upload. If this happens it is also possible that the sweeper that moves these files into place will move a partially written form of the file into the archive; a Bad Thing (tm).
Currently we depend on Resumable to generate the "unique" identifier for the upload. It seems a better solution to have the client
preparean upload, and have the WS generate a (truly) unique ID in response. This way we can guarantee no collisions between uploads and prevent a second upload with the same ID from starting if the file has already been successfully staged.