I personally think uploading massive files is not usually desirable. Better would be an easy way to chunk it and upload and have the server put the file back together, which would increase reliability.
Tus is over-complicated IMO. Why require the server to store state that's already available on the file system? Just support PATCH requests with an offset parameter. If an upload fails partway through, HEAD the file to see how much got copied to disk and resume from there.
