On 1/4/2018 10:21 AM, Stephane Bortzmeyer wrote: > I've seen a lot of hand-waving how how a rootless system may work, but > never a detailed technical analysis, covering all the cases, specially > the inconvenient ones (such as .home). > > Same issues for alternative (non-DNS) naming systems: most sweep the > problem of consistency under the carpet. (GNUnet is one of the few who > honestly declare that they don't try to have unique names.) Yes. The key to decentralization is support for non-uniqueness. It requires one level of indirection from "mnemonic names" such as "example.com" to "decentralized unique identifiers", typically derived from public keys. But that still does not work if the resolution system gives a single answer, as in "this is who the system believes is example.com". The client would ask for the key, see that it doesn't match, and be stuck. So the resolution system needs to answer "there are multiple entities who claim to be example.com, here is a list", and the client would then have to check which one has the right key. And then of course there is the question of first contact, when the client does not know yet the "right key" for example.com. This could be done, but it would require a huge amount of work -- boiling the ocean comes to mind. -- Christian Huitema