[dragonwriter]

> There is no way - not even a theoretical way - to allow blocking of illegal content (for any definition of illegal) that won't allow for blocking of any other arbitrary content.

Well, there is: take the person or body that ultimately determines whether content is illegal, and have them review each request and proposed response and decide whether to allow the content through to the requester.

For slightly better scalability, have that body review all content outside of any request-response cycle before it can be published and sign any approved content, then block any content they haven’t signed.

Somewhat more generally, as long as the specific blocking methodology is itself part of the definition of what content is legal, any blocking method can meet the standard of “allows blocking illegal content without allowing blocking of other content”, since any content blocked by the method is, ipso facto, illegal.