It's simply a "mapping" service and no more and hardly differs from a DHT w/ replication.
Julian Cain On Dec 10, 2010, at 12:36 PM, Michael Rogers <[email protected]> wrote: > On 10/12/10 14:14, Stephane Bortzmeyer wrote: >> DNS is not a discovery service (search engines are better for >> that). It is a naming and mapping service, with stable names. > > I'd argue that in its current form, DNS is both a discovery service and > a stable naming service. If someone tells me "I think you'd enjoy > wasting countless hours of your life on slashdot.org", and I type > slashdot.org into my browser, DNS is being used as a discovery service, > and that's the function I'm saying could be replaced with search engines. > > On the other hand if someone sends me a link to slashdot.org and I click > on it, DNS is being used as a stable naming service, and that's the > function I'm saying could be replaced with a decentralised > public-key-to-IP-address mapping service (where each mapping would be > timestamped and signed with the corresponding private key). > > The reason for separating these two functions is that a discovery > service requires human-memorable names, and therefore requires trust in > the party who provides the mapping, whereas a stable naming service > doesn't require any trust (because everything's cryptographically > verifiable) and can therefore be implemented in a decentralised way. > > Fortunately the discovery function of DNS is only used rarely, so I'm > arguing that it's alright to ask the user to make a judgement call in > such cases, whereas the stable naming function is used very frequently > and needs to be automatic. > >> The fact that you type names or not is irrelevant. To send this >> message, I did not type '[email protected]', I just replied >> to your email. Nevertheless, I relied on the DNS to be sure it is sent >> to the proper machine, without fuzzy matching and without long >> explorations of possible results (and subsequent questions to the >> sender). > > Yes, absolutely, search engines should only be used in rare cases like > word-of-mouth recommendations where DNS is currently used as a discovery > service - in all other cases we should use a decentralised system that > doesn't require human-memorable names. For example, this list might be > called something like > [email protected], and both of us > would just reply to that address without ever trying to memorise it. > > That's why I asked how often you actually type a domain name. In any > situation (such as replying to an email) where you don't manually enter > the domain, it doesn't need to be a human-memorable name. > > 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
