I agree with some people here who are a little skeptical about the proposal. It introduces a new trusted third party, which is opposed to Nostr's spirit.

Have you guys also considered an approach in which the user generates a key pair (account) in every new app and maps the accounts together as belonging to one identity? Losing one key can be handled by making it invalid through a voting process achieving a certain quorum. In a similar manner, the user can add new accounts to his identity by voting for it with existing accounts and reaching the required quorum.

I imagine the new protocol can define events of new kinds for account mapping requests, approvals, and rejections.

There are some open issues with this approach, for sure. How we can minimize attack vectors and handle the fuzzy state inherent to Nostr? But I think it is worth exploring such an approach too. With this approach even browser extensions aren't needed, replacing trusted third parties through a consensus protocol.

Do I miss something fundamentally wrong here? I'm looking forward to your thoughts on this.