One big difference though is that the users on one server could communicate with users on another server. In the world of IRC that would mean that users from Freenode and EFnet could communicate with each other in the same channel. Pretty awesome.
- Do I necessarily want my chat to be intermingled with #python on EFNnet, OFTC, etc.?
- How would you prevent new users from auto-subscribing to a spam server? Do you just ban the spammy server? Would every server operator need to manage such a ban individually?
Handling spam and abuse in general is a huge issue for Matrix which we're putting massive focus on at the moment (in fact, it's why we're in SF at the Decentralized Web Summit currently). The whole decentralized web movement has a crisis on the horizon if there isn't a decentralized identity/reputation/abuse system of some kind to track miscreants and help admins and users mitigate spam.
Edit: in other words, right now a server admin would have to manually blacklist bad servers. In future there'd be a decentralised data structure of some kind (stellar ledger, blockchain, matrix DAG, IPFS DAG or whatever) tracking the greylist of servers/users/gateways to help decide who you want to talk to.
There was only one network and any server could join it. But you could set up malicious servers can cause havoc on the network. Eventually all the Hubs but one closed their doors so only authenticated servers could be added. The remaining holdout was eris.berkeley.edu
And so the network split into the Eris-Free network (or EFnet) with the remaining servers connected to eris forming the Anarchy network (or Anet). Anet died out reasonably quickly.
There have been a few more splits over the years, until you reach the current situation, with around 6 major networks and hundreds of smaller networks.