A bundle event is already part of the packaging. It references files by their MMR root hash.
By using mmr, relays can be sure the publisher isn't lying about the number of chunks and that a chunk is exactly at x position. The chunk size is fixed. So relays can choose to block a chunk if its part of a huge file.
Imho blossom isn't needed. Although relays specialized on storing big events could rise, the napp files tend to be so small that we write them to the publisher's outbox relays.
Nothing is "needed", I'm just stating what is better.
Blossom is better because it's more efficient, it will work every time and not hit any barriers, it is much simpler to implement from the publisher side and from the reader side, it's also nicer to use other people's infrastructures the way they were designed to be used.
If we want this Napp model to become a standard (and we need that if we are to solve the problems I'm pointing) then these are all very valuable properties.
I won't argue with the protocol creator: you win hahah
I will just add that the choice of using relays was also based on a longer term goal of offering a local backup to all the user data and it will be easier to implement if everything is expected to be a Nostr event.
