Yeah I'm envisioning a setup like this:
- Users upload media to hosts like nostr.build/nostrimg/etc who act as persistent bittorrent web seeds
- Users who then download that media from the web seed can then act as seeders themselves if they're on desktop, their client supports it, and they opt in.
This way we can try to find some middleground compromise that combines the persistence of centralized media hosting with the (theoretically) more evenly distributed bandwidth load of full P2P.
I’d like to see some performance testing in clients for time-to-load images as examples. A direct host/source is going to be faster I imagine.
I suspect torrents may be too slow for normal use and may need to be the fallback - and clients are incentivised to make it fast for best UX. Unsure how this impacts the economics of it.
A benefit from torrents however is when something is older or less cached/recent, you have a common addressable was to query across a peer network for who has this. That’s where I think it adds the most value.
