I don't see why right to be forgotten is fundamentally incompatible with Fediverse (or other social web 4.0, idk the name)? For example, there could be a central server where you publish some credentials along with a request to be forgotten, and all instances must (if they want to comply with the law) check there from time to time (this would happen automatically of course) and erase associated content. Or you could ask to be forgotten to your instances and it would message all relevant instances of your wish to be forgotten, which then is their own duty to comply. It doesn't seem like an insurmountable issue.
I think the same thing, but it only takes a temporary instance whose goal is just taking your data and selling it to a broker then shutting down to defeat it
Yes, but in any case the data is already public, or at least most of it is. I think if we truly desire for privacy of posts then the SWF/Fediverse should be looking at a different access model, and maybe some encryption methods (for example you could whitelist who can decrypt your data).
There's also the legal approach: keep data public but forbid usage for unauthorized purposes. Then at least the big public companies might not put their hands in it (at least not directly). But it could still be leaked and widely used by nefarious or shady actors.
In any case I think a more well thought out or just better informed distinction between public posts and private ones would be very useful (it appears currently in Mastodon you need to have a 'locked account' to post followers-only posts; I don't know if data goes to other servers for private posts?). If you post something anyone can see on the internet, there's a significant limit to how much the law or any technical solution could protect you. The right solution for sensitive data on the internet is to make sure only a select few can see it, for now.
(I believe this problem is can actually be solved using something like a Web of Trust (but more formalized/automated than WoT is currently): you'd trust accounts and then you can build a graph where trust is a function of distance (trusted node distance). This web of trust could be useful for a number of distributed protocols. The only catch is it requires much more user interaction and education.)
