Concurrent upload and chunk concatenation

[izissise]

Right now if there is concurrent uploads with the same chunk size, there is no way to correctly choose chunks to reconstruct the file

Some workaround that could be implemented:

• randomize chunk size
• if the chunk size is the same for two file in the same client, serialize the upload
• hash file and put hash in meta, file can hashed will being send (to avoid doubl read) and the meta upload connection stays open until the end to push meta from the end
• find a way for nginx to send back the chunks filenames and put the list in the metadata

[izissise]

I started to implement hashed upload here https://github.com/izissise/nginx-explorer/tree/hash-upload Sadly firefox doesn't support upload from readable stream, that make it inneficient because we would need to read its chunk twice first to hash it and then to send it.

https://bugzilla.mozilla.org/show_bug.cgi?id=1387483