On Thu, 2010-12-09 at 03:35 +0000, Michael Rogers wrote: > On 06/12/10 21:31, Julian Cain wrote: > > This doesn't fix the problem which is the US federal government. > > The problem isn't any particular government. The problem is that any > global, hierarchical namespace requires global, hierarchical management. > Even if the method of distributing .p2p DNS records were completely > decentralised, *somebody* would still have to decide who owned which > domains within the .p2p namespace, and those owners would then have to > decide who owned which subdomains. > > > The only way to fix this permanently is to take back control, build a new > > DNS > > infrastructure, and deploy it. > > It's worse than that - the only way to manage without hierarchical > control is to give up one of two things: global agreement about who owns > which name, or the ability to prevent name-squatting. > > For the sake of argument, let's imagine that we give up global > agreement. Perhaps in the future a DNS lookup might work something like > a web search works today: I go to a semi-trusted broker and ask that > broker to resolve a name for me. The broker returns one or more results. > If two or more owners claim the same name, the broker either selects > one, or ranks them. I have the option of asking another broker the same > question and comparing the results. Anyone can switch brokers at any > time. At any given moment there's an unstable consensus among users > about which brokers are usually reliable, and an unstable consensus > among brokers about which names correspond to which owners.
The following work tries to do exactly that. It's called SocialDNS and it works by allowing a user to create a DNS mapping (aliceblog.sdns -> 172.31.34.15) this mapping is then shared through a social p2p network where social relationships are mapped to encrypted P2P links. These mappings can then be propagated through the social p2p graph the same way a file gets replicated through a gnutella p2p graph as more people discover and download it. Anyone can pick any DNS mapping (ending with .sdns), collisions are solved through a ranking system thus the most popular name is chosen similar to a web search where the most popular website is put at the top. This system is currently used to provide a decentralized naming mechanism for SocialVPN, which is a P2P VPN that uses XMPP to automatically create a VPN with friends. This is the link to the paper http://www.acis.ufl.edu/~ptony82/papers/collaboratecom10.pdf This is the link to the talk http://www.acis.ufl.edu/~ptony82/talks/collaboratecom10.pdf This is the link to the software http://socialvpn.org -- Pierre > > Since the lookup process is a bit fuzzy and requires human judgement, > it's only be used in situations that absolutely require a > human-memorable name, such as word-of-mouth recommendations. Most of the > time - in links, bookmarks, configuration files, email headers - > machines are referred to by unambiguous, verifiable, but unmemorable > names, such as the hashes of their public keys, which can be resolved to > IP addresses via decentralised infrastructure without requiring > centralised control. > > In other words: who needs DNS if we have reliable search engines and > HIP? How many domain names do you type per day anyway? > > Cheers, > Michael > _______________________________________________ > p2p-hackers mailing list > [email protected] > http://lists.zooko.com/mailman/listinfo/p2p-hackers _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
