Probably because until the document is saved, it doesn’t exist on the server. How will the server accept a file for a document that doesn’t exist?
Obviously workarounds are possible, but it would be complicated. Various choices present themselves:
- don’t do the actual upload until the ‘save’ button is pressed (and ‘save’ first saves, then uploads). This may break user expectations that the upload occurs at the time, and that ‘save’ should be a quick operation.
- uploaded files are stored against temporary files which are deleted after some time elapses. Probably a bit of a minefield to actually implement.
All would, I’m sure, require a fair bit of work - which is probably why they didn’t do it that way.