The fediverse works by propagating statuses a set of standard protocols: https://en.wikipedia.org/wiki/OStatus https://en.wikipedia.org/wiki/ActivityPub
Typical Mastodon instances have one or more frontends (Ruby on Rails) backed by a PostgreSQL database (possibly replicated). Several cloud providers offer managed PostgreSQL instances which mitigate the SPOF of using an RDBMS backend.
Accounts of users are bound to the instance where they were created, but recently Mastodon added a data export function that could be used to move an account to a different instance. I doubt you'd retain your followers in this case.
Unboundling user identity from instances sounds like a difficult researh problem, given that users can upload considerable amounts of data.
And it'll work with other mastodon instances.
Mastodon isn't hotmail or gmail it's an MTA+MDA, anyone can setup those on their own servers and become reachable by email...
Note that this also makes it hard to deal with spam and DoS attacks. As adoption grows, instances might voluntarily adopt protocols for sharing abuse data.
No.
> How does Mastadon handle database (multiple masters, etc)?
Each node has a partial knowledge of the network.
> Genuinely curious if this is truly decentralized platform or just a poor man’s open source social networking app with some RSS type of functionality sprinkled on top of it, sort of like Wordpress in a sense in terms of hosting it.
Neither. It's federated, like emails. Users still have to subscribe to a server, but anyone can run a server.
Mastodon depends on postgres; whether your postgres is multi-master or such is entirely up to you, the instance admin. Hosting it is much like hosting any other modern rails app.
Some admins use rds, some use postgres-in-docker on a single machine, etc etc. I'm sure some have more fascinating setups there.
However, that's wholly outside of what mastodon claims to do or be; it's decentralized and federated in more meaningful ways; there is no central authority listing users or allowing logins.
If your server goes down for some reason, much like email, you'll lose the ability to log in and post, but anyone on a different server will be unaffected.
I don't believe there's any particularly special DB replication features in the reference server implementation, but there's nothing inherent in the system that would prevent someone writing a customized durable/distributed version of the server.