[imtringued]

Federated systems are the middle ground between full centralisation (one provider) and decentralisation (everyone is a provider). They allow users to choose between the two depending on what they actually want.

With decentralised systems you are forced to run all computation and store all data yourself. The average user isn't going to buy their own server and run it 24/7. That's incredibly wasteful. Even in the best case scenario they will just get an "urbit server" from AWS or somewhere else.

The problem with centralised services isn't that they are centralised. The problem is that you are locked-in to a specific provider. You can't send messages to your facebook friends from your freedom respecting gnu social server or whatever. You have to install a dozen apps (facebook, whatsapp, telegram, etc) to talk with all your friends.

[chc4]

That's pretty much the end goal of Urbit, though. It doesn't want everyone running a server in their closet, just a server in general which includes on AWS. The point is that it should be easy enough to control that your mother can do it: pay eth, click a button, get urbit running in the cloud should be the final flow.

Having a common platform, all using the same RPC and identity and network layer, allows for the removal of lock-in across different Urbit apps.

And since it's your own server, you /can/ send to Facebook friends from your Urbit clone just by hitting the API endpoints. Migrating to an Urbit clone of a centralized service should be gradual to fight back against network effects: first make a UI frontend for the service from Urbit, then make it mirror content on Urbit back, and gradually transition over. GNU Social does this with Twitter, but kinda badly, and can be locked out from one central API token because everyone has to go though their node.