Oddbean

The very fact that we insist on identity sovereignty puts us at a disadvantage vs. say Bluesky where people pretend that they own their DID, but they don't. So if we can't avoid the onboarding cost of managing a root key, at the very least we owe it to ourselves to minimise that UX cost to the least necessary.

So you’re not using the priv key to sign any and every thing? For nostr, I was looking into a hardware signer device that would be able to have a whole nsecbunker running on it. But it seems like that doesn’t exist anymore

Not even some things, the only thing we sign (yet) is just a very short lived authentication token, which you send it to your homeserver and get a good old cookie (session) in return, and you use that going forward. In fact we also have a 3rd party authorization spec(and working code), so you can allow a web app to obtain that session without uploading the keys to that 3rd party app. We will properly go even further and make your root key only used once in the beginning, then use a delegated key for logging in when you get signed out... but we haven't went that far yet. For now, you use your key as often as you sign in to Gmail... very rarely

DHTs had a horrible PR past decade. Some like Fiatjaf and others only experienced them trhough and libp2p which is rightfully shit. Others who know better think that it is not viable to use a DHT as a database, missing the entire point, which is that a temporary censorship resistant routing table is better tha no censorship resistant table at all. I just went on to validate if it works, knowing that if it doesn't, nothing else will and we can just stop lapping. Sorry, but we will never bootstrap a million nodes of anything ever again. Finally some people are skeptic of DHTs vulnerability to Sybil attacks, and I am writing something about that to end this FUD once and for all, but even if they are right, the question remain, compared to what? Because there literally is no second best within the same order of magnitude even.