Your comment does not make sense because it fundamentally says that the spec needs to define the DHT so that no 10 name systems exist. But that is simply nonsen

Your comment does not make sense because it fundamentally says that the spec needs to define the DHT so that no 10 name systems exist. But that is simply nonsense. If you implement a new DNS stub resolver that follows the record wire formats but uses different root servers it will be effectively not the "same" name system as vanilla DNS. It will have different records under names and certainly different security properties. That is the whole point of hyper-local root for DNS. I am simply implying that you are wrong in that we have to specify the underlying layer. Anyone can (should be) able to use the spec an implement it. It could be done on top of Bittorrent (not Kademlia as that is the protocol, not the network!), IPFS (the DHT they use) or any other DHT. With the respective implicatoins on security. It will still be conforming to the spec.