On 18 Sep 2015, at 9:54, Alec Muffett wrote:
>> On Sep 18, 2015, at 14:16, George Michaelson <g...@algebras.org> wrote: >> >> My private comment bears repeating in public. >> >> DOMAIN names is about the property of domains. Domains are encompassing, >> set-theory/venn-diagram style. A domain and a prefix are analogous concepts. >> One is expressed syntactically somehow, the other is a mathematical property >> of bounding in a number field but they have the same basic behaviour. >> >> the UK domain order in coloured book mails obeyed this property: it just >> used reverse semantics to the ARPA model. >> >> XXXXXXXX.onion is *not* a domain name inside the .onion part: as I >> understand it, the value is a hash, or other function which has no nesting >> properties expressed syntactically. > > Hi, my name's Alec, I work for Facebook and lead the engineering team for > Facebook over Tor. This reminds me of the time I set down with a collection of people who would later turn into NZNOG, at a Uniforum meeting in Taupo. Since we were sitting in a circle, it seemed only natural to start things off with "My name is Joe, and I work for an ISP". Everybody else without missing a beat replied with the twelve-step "Hi Joe". We had a moment. Hi Alec! > You are certainly correct that the label immediately left of ".onion" is a > hash, and functions not unlike a layer-3 address; however, there may be other > labels leftwards of the hash, under (to some extent) other administrative > control. I think that we are all guilty from time to time of trying to form elegant, general descriptions of things that are not actually elegant, or useful to generalise. The DNS is frequently described has having three core concepts: (a) the servers and the wire-format protocols that they talk, (b) the data model (resource records, etc) and (c) the namespace. (a) provides the infrastructure for (b) to be retrieved using a key from (c). There are other name resolution protocols that are not the DNS, but which use similar namespaces to (c) and perhaps similar (b) but different (a). Pertinent examples are multicast DNS and Onion/tor, and (arguably) the localhost "protocol" that simply maps the name localhost to the addresses 127.0.0.1 and ::1. The ugliness all rotates around the pragmatic decision to use the right-most label in a name as a resolution protocol selector. We can complain about that all we like, but reality is that we're going to have a hard time pushing those cats back into the bag. At the very least there will be injuries and bleeding, and you know the cats aren't going to like it. It would be lovely if nobody had ever used the right-most label like this, and instead there was a standard and accepted way to specify a resolution protocol in a URI, and everywhere else that a name is used. But there isn't. Also, running code, etc. Whether or not we should call an onion or mdns name a "domain name" or something else is just a detail. I don't think agreeing on the answer is going to solve any of the problems that we actually have. Joe
signature.asc
Description: OpenPGP digital signature
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
