Now that we’re in the process of implementing multi-part uploads, I was thinking about that a bit more, and this specific question stuck in my head:
Going back to first principles, the reason why we have to design the upload protocol in the first place is that there’s no HTTP RFC for this problem. But if that’s the case, why solve this problem only for ourselves, when the idea of RS is to just use a subset of other open protocols, with minimal sprinkling on top to tie everything together.
So, the question is: wouldn’t it make more sense to submit this as a standalone multi-part upload protocol and only reference it as an optional add-on for servers and clients (using our Webfinger add-on discovery) in the RS spec?
Same goes for the implementation then, of course. We could make the pure upload functionality its own library that any other client, not just RS ones, could import and use.