It’s not absolutely safe against bots and sockpuppets, but it surely makes it more expensive than even a $10/account membership.

I think, sadly, that either sending in your national ID or paying $10 would be unacceptable to so many people that it would make it a lonesome failure of an experiment. I'm on your side about the idea, but I think people would just take the path of least resistance and create their sockpuppets on some other instance, and your main accomplishment would be driving away legitimate users.

PIxelfed is still just supporting ActivityPub. I’m talking about multi-protocol communication. A smart client should be able to let you communicate with Lemmy communities, subreddits, Facebook groups and all types of different platforms from a single unified interface. There are plenty of people that think this is something undesirable (like everyone that wants instances to block Threads), but I’d argue that building these integrations with closed platforms would eventually destroy them because they would lose the monopoly on network effects.

I get it. Aren't there projects that are working on that? Friendica and Emissary? Adding integrations with closed-source networks to those isn't too hard. At that point, it's not its own web app anymore, though, more akin to an email program. It's a good idea but it's different than what I had in mind. You will also have to deal with API limits or terms of service and legal issues, once you start looping in the closed-source networks.

No, but you could have a web server that responds to multiple domains. Ideally, the server listening and responding to the AP requests should be able to work with multiple “virtual servers”, instead of having to have only one instance == one domain that we today. AFAIK, only Takahe does this for microblogging.

Yes, that part's not overly hard. I'm already doing virtual servers for ponder.cat and rss.ponder.cat, to run them both on the same VPS, and I'll probably add more virtual servers for development of frontend tweaks if I keep going with Lemmy. Some of the ideas I had in mind for hackable frontends involved wildcard virtual servers to serve people custom "instance" sites off a subdomain that's different from the actual actor ID instance name.

What I'm saying is that if someone's actor ID from the POV of the rest of the Fediverse is still https://ponder.cat/u/rglullis, and ponder.cat goes down, nothing that either ponder.cat or any new instance can do, can "catch" requests that are being directed to that actor ID. You have to make the actor ID either https://rglullis.com/u/rglullis or https://rglullis.sometrustedthirdparty.com/u/rglullis from the beginning, and arrange for ponder.cat to be handling any traffic for those domains, so that you can switch away from the ponder.cat instance later on if you want to.

Of course, you can tell people that they can either have a ponder.cat user, or a rglullis.com user if they want to buy their own domain for their user, and they can have an actor that will be transferrable from ponder.cat to any other Lemmy server that supports the feature. It wouldn't work with current Lemmy, but in theory it could be made to work, if someone were willing to make the right Lemmy changes. It would be tough but it might be worth it.

Overall I think it might be better to address the same issue at the protocol level as some other federated social media networks do, so you're not introducing crazy new requirements on both the server and user experience side in order for people to be able to transfer their users later.