Interesting. Maybe the big idea is not the relay but the ‘atomic event’ (signed event). It either exists or doesn’t exist.
Discussion
Both in unison. They solve for different problems.
I've been thinking about this too. The events can exist without the relays, technically, but the relays create the landscape. If the relays disappear, we're back to centralized servers, and there's no incentive for event interoperability.
Yes, relays multiple the events and spread them across many different, connected machines.
We spend a lot of time on data visualization. Maybe we should also visualize the relay system, nostr:nprofile1qqsdcnxssmxheed3sv4d7n7azggj3xyq6tr799dukrngfsq6emnhcpspz9mhxue69uhkummnw3ezuamfdejj7qgmwaehxw309a6xsetxdaex2um59ehx7um5wgcjucm0d5hsz9nhwden5te0dp5hxapwdehhxarj9ekxzmny9u78m52f , and calculate if the event storage and transmission is getting more or less centralized.
"Landscape" immediately sounds like something worth mapping and navigating by.
Perhaps we could cluster events by relay, or assign each relay a color, or give each event a number indicating how many relays hold a copy.
Even more far-out, a visualization could use relays as a third axis. A 2D plane shows the events and how they relate to each other on one relay. Look at it from the side in 3D, and there are many planes stacked atop one another, each one representing a relay.
It would be interesting to verify this. Nostr likely does have a power law distribution of users to relays. Preferential attatchment describes that when new users join a network they will join in a biased way, connecting to their friends and those that already have an established presence. Considering that Outbox isn't widely used, that many of the onboarding clients use their own relays, and that general users would be less likely to change their relay settings unless told to do so, I'd expect that we're still very centralized. Key factor - microblogging incentivizes following users, not content.