bramcordie / blobload Goto Github PK

Tokens are now generated on the client side based on the file name concatenated with the size in bytes.
This is wrong and was done for practical testing reasons.

The upload service should be handing out the tokens. When the client tries to upload a new file he should calculate a checksum of the file. Because the client calculates the checksum it can only be trusted as a reference for resumed uploads. After the client hands you a checksum you give him a token he can use to upload his file.

After an upload is completed the service can hand the same token to your application for future reference.

When multiple instances exist of the upload of the same file (e.g. in 2 different browsers), duplicate slices are sent to the server, leading to improper file buildup on the server. Identification of already received slices is necessary before the slices are added to the buffer.