Phony "federated apps" are mostly "fragile self-hosted with linkrot, convenience, and pseudo robustness".
Robust federation works as a distributed overlay network and doesn't require any leader. The irreducible issues become:
0. "Which systems should store what data, i.e., blobs (files), entities, and entity sequences?"
1. "How many copies should there be of 0.)?"
(1.5. "What will keep scrubbing 0.) for integrity and duplicating 1.) below a given threshold?")
2. "Where should functions against 0.-1. run?" #
3. "How many copies of 2.) should execute?" #
4. "How should the operators of persistent systems recoup the microtransactional costs of compute, storage, and networking of 0.-4.)?" (Client has a pool of credits purchased through some crypto means used to rent storage capacity, net transit, and processing of media, metadata, and code #.)
5. "How many copies of next nodes and node paths do you maintain?"
6. "Which nodes should this node remain connected to?"
7. "How many fixed default nodes can be run around the world to always seed a node's initial network topology?"
8. "How much anti-correlation traffic should fill the encrypted link when there is no traffic?" (Otherwise, it becomes very easy to poison and unmask overlay networks.)
# If the platform has a serverless function concept, where it's unknown where it will run until it does.
Robust federation works as a distributed overlay network and doesn't require any leader. The irreducible issues become:
0. "Which systems should store what data, i.e., blobs (files), entities, and entity sequences?"
1. "How many copies should there be of 0.)?"
(1.5. "What will keep scrubbing 0.) for integrity and duplicating 1.) below a given threshold?")
2. "Where should functions against 0.-1. run?" #
3. "How many copies of 2.) should execute?" #
4. "How should the operators of persistent systems recoup the microtransactional costs of compute, storage, and networking of 0.-4.)?" (Client has a pool of credits purchased through some crypto means used to rent storage capacity, net transit, and processing of media, metadata, and code #.)
5. "How many copies of next nodes and node paths do you maintain?"
6. "Which nodes should this node remain connected to?"
7. "How many fixed default nodes can be run around the world to always seed a node's initial network topology?"
8. "How much anti-correlation traffic should fill the encrypted link when there is no traffic?" (Otherwise, it becomes very easy to poison and unmask overlay networks.)
# If the platform has a serverless function concept, where it's unknown where it will run until it does.