I've also come to the conclusion that a tightly designed subscription service is the way to go. Cheap really can be better than "free" if done right.
Even in a federated system, you can be blacklisted although it does take more coordination and work.
i2p and writing to the blockchain are an attempt to deal with that through permanence, but those are not without their own (serious) problems.
Payment processors underpin ad systems and they have strong leverage to pressure ad buyers and can pull your ability to make those sales. That's on top of the advertisers themselves having strong positions on what kind of content they want to advertise beside.
Everyone has to pay for servers somehow. Especially at scale. And doing that without payment processors is difficult. Crypto has not proven itself to be something consumers will use.
In all reality, the solution to as much free speech as possible on a social platform is to limit reach. If people want to broadcast to millions or even billions, then of course that will come with limitations and restrictions. Everyone has to balance the varied interests required to achieve scale. Limiting individual reach means more potential freedom for users.
I'm not against subscriptions per se, but I do think a one time entry cost is really all that's needed to achieve many of the desired effects. I'm probably in the minority as someone who'd rather pay $10 one time to enter a community once than $1-2/month to maintain my participation, though. I'm just personally tired of feeling like I'm paying a tax to construct something that may one day be good, rather than buying into a decently polished product upfront.
And even if it works, there will still be carry-over of many of the problems we've seen with centralized social networks.
It's oxymoronic. Our communication was developed in highly developed hierarchies for a reason: continual deception, deviance, anarchism, perversion, subversion always operating in conflict and in contrary to hierarchies.
Language is not self-organizing, signaling is not self-learning it self-regulating. The web opened the already existing pandora's box of Shannon's admittedly non-psychologically relevant info theory and went bust at scale.
Something like OAuth means that you can give different levels of private data to different actors, based on what perms they request.
Then you just have whoever is holding your data anyway (it's gotta live somewhere) also handle the OAuth keys. That's how the Bluesky PDS system works, basically.
Now, there is an issue with blanket requesting/granting of perms (which an end user isn't necessarily going to know about), but IMO all that's missing from the Bluesky-style system is to have a way to reject individual OAuth grants (for example, making it so Bluesky doesn't have access to reading my likes, but it does have access to writing to my likes).
If I have 100 followers on 100 different nodes, that means each node has access to (and holds on to) some portion of my data by way of those followers.
In a centralized system, a user having total control over their data (and the ability to delete it) is more feasible. I'm not saying modern systems are great about this, GDPR was necessary to force their hands, but federation makes it more technically difficult.
I don't really see my posts remaining in people's RSS readers after deletion as a problem. It's a fundamental property of information distribution as far as I am concerned.
But even then, that means there's resistance, but that's not the same as things being technically impossible. In federated systems, a delete is not a delete. It can't be because there's no way to confirm deletion on nodes you can't control.
And I understand your perspective as a realist on deletion generally, but that's not most social media users understanding when they're told they can control their own data, which is a common selling point of federation.
A centralized system which is properly incentivized to completely wipe all data associated with an account will be able do so, but a federated system can't.