But that doesn't solve the problem, as the "political wars" the person I was responding to described all work in the world of increasingly tight laws; you have to avoid being a target in the first place by correctly navigating the space of solutions for how content is distributed in the first place. Take the Apple iPhone example: if Apple is forced to remove an app from their app store by China, it doesn't matter if the user can install it on the side... and somehow China has not managed to force companies to not provide that functionality: Apple going out of their way to centralize both app distribution and even API access (VPN functionality, as a specific example, is hidden behind an entitlement that they don't give to casual developers) is work they chose to do that directly enables the ability to even make "legal requests" quickly and easily.

I maintain the moral equivalent of this for platforms like Twitter and YouTube is to stop conflating their centralized recommendation systems (which should be considered "their speech") with the functionality to publish at all (which should be considered "someone else's speech"): it is going to be way less controversial to stop recommending something that someone else likes if it is still accessible to people who know about it (as they externally discovered the content or author and were directly linked to it/them), and it is also going to be way less controversial to allow people to publish something that someone else doesn't like to their own audience if it isn't being actively recommended to third parties. These recommendation systems have started to conflate "being able to say what you want" with "being able to be granted a large audience" so well that the feature set for moderation fails to separate them, and that's bad for everyone: there is a reason why we talk about "the right to free speech" instead of "the right to be heard".

> If a website does not honour legal request, it'll get blocked

If that is happening, then the architecture of the internet needs to take more evolutionary steps.

If the state can censor the internet, there's nothing 'inter' about it. It's just a network, subject to the whims of its flailing predecessor.

> If that is happening

Yes, that is happening for many years. China being the most known state, but AFAIK even well-recognized first-world democratic countries like UK block some websites.

> then the architecture of the internet needs to take more evolutionary steps.

Adoption of IPv6 is a good example that architecture of the Internet is pretty much set in stone at this moment. We can put more layers on top of it, like Tor network, but underlying protocols are still IPv4/IPv6 with enough meta-information to allow efficient blocking of protocols or resources.

I have some hopes that new TLS standards with hostname encryption (ECH) along with CDN networks will make blocking impossible. But even that is easily circumvented with government MITM. Kazakhstan already deployed all the necessary hardware and did some successful tests on scale. Browsers blocked its root certificate, but will they block (imaginary) China root certificate, losing 1.5B users?