Hacker News Reader

Comment by wmf | Hacker News Reader

Decentralized DNS has been discussed, but it runs afoul of Zooko's triangle. In such a system, human-readable names would end up resolving to different IPs for different clients, which breaks everything. You can reduce but not eliminate the breakage using something like a Levien trust metric.

View on HN

Distribution of update authority doesn't require decentralization of the namespace itself. As long as the independent authorities can agree on what constitutes a valid update (say by requiring the domain-owner's signature), they'll stay consistent. Meanwhile a corrupted authority will only affect its own clients.

In that case I would be worried about having a quorum of authorities located in the USA and all "corrupted" by law, thus outvoting the non-censored authorities.

I'm not talking about any sort of voting, just restricting the inconsistencies only to answers from the corrupted authorities. If cryptome.org redirects to thoughtcrime.dhs.gov for DNS servers using US roots, it has the same localized effect as any current national dns tampering. All uncorrupted authorities would all continue to agree that the domain hadn't been deregistered because they saw no signed update from cryptome.org. (And of course an 'authority' can even be your local upgraded DNS server if you care enough about having uncensored DNS)

edit: Actually I guess I'm talking about a quorum of 1 for legacy clients (nationality of chosen root servers), and updated authorities/clients relying on the total quorum that's easily established before the domain becomes "of interest".

Reply on news.ycombinator.com