If you think through the implications of my proposal you'll see that it is only invalidating hostnames in these ICANN-giveaway new "TLDs". Traditional domains would remain unchanged. Trustees of traditional TLDs would still continue their corruption (as was narrowly avoided for .org a few months ago), but that's probably inevitable. Meanwhile we'd preserve the root namespace for the adoption of better technologies (eg .onion).

And honestly we need better mechanics for machine-intended references regardless. It's ridiculous that you resolve and load a webpage, only for that webpage to require you to resolve a bunch more human readable (ie non-decentralized) names for loading subresources. For example, going to a bookmarked page shouldn't result in any DNS queries for human readable names.