i think that deep discussion over whether .external is the right exit gateway from dns naming is premature, and that we should first decide whether a single exit gateway is preferred, and whether IANA should craft a registry of external-to-the-dns uses of the internet name space. i am in favour of a single exit, and a registry, with a pointer to an RFC for any second-level term under that exit label (placeholdered as .external but requiring discussion later if we get that far.)
i'd hope to see apple's mdns work migrate to this new system, as .zeroconf.external or even .bonjour.external, and the /etc/hosts lookup mechanism migrate to .hosts.external, and sun's yellow pages system migrate to .nis.external, and so on. i would argue, by the way, that "onion" is a kind of technology, onion routing, of which Tor is the first and best-known but not the last. so, i'll prefer .tor.external over .onion.external. this whole thread is in the area of "dns presentation layer" which is quite ill-defined. users and apps currently mix in non-dns strings which can look fully qualified (have some dots in the middle) with dns strings, and we're counting on nature to avoid collisions for us. ideally we'll define an exit gateway such that non-dns names can be both expressed to a user through an application, and entered by a user into an application, for which collisions are prevented by design and artifice. the application should not need to know anything about these naming policies, because the operating system and/or name lookup libraries should just work. an alternative to a reserved top-level domain as an exit gateway for the dns presentation layer would be something along the lines of what IDN does, prepend some characters and dashes to a name in order to ensure that it cannot conflict. so, instead of .onion, something like .xq--tor. it's my view that this is uglier and also has bad human factors engineering (hard to read and hard to type) but it should be mentioned in passing as a road deliberately not taken, to inform historians in the future. Andrew Sullivan wrote: > On Fri, Jul 17, 2015 at 12:51:05AM -0700, Paul Vixie wrote: >> yes, but not with .ALT, which is a politically desirable gTLD name, and >> which allows the connotation of "alternate DNS". i suggested .EXTERNAL >> because nobody will ever want it as a gTLD and because its connotation >> is unambiguously "not DNS". > > Do you have evidence that alt is in fact a desirable gTLD name? as you know, proof is impossible. i'm working backward from .xyz and also this: https://en.wikipedia.org/wiki/Alt.*_hierarchy > I > note that we recently went through an empirical proof of desirability > of such names, and at the very least it was not worth $185k to anyone. as you know, the current gTLD round will not be the last. > I guess you might be drawing a distinction between "economically > desirable" and "politcally desirable", but I'm not sure how to measure > the latter. there are rock musicians, medicinal practitioners, historians, and energy researchers working right now today all over the world who call their brand of work "alternative". so, consider whether "artistically desirable" is a third category, and whether there can't be a fourth. > One problem with "external" is that it's long. Part of the use case, > recall, is that people want things they can bang into any old domain > name slot in any existing application, and it'll fit. Using up 3 > octets isn't so bad. Using up 8 seems wasteful. Would "ext" do the > job for you? compared to alt, yes. note that .external is long on purpose-- to avoid conflict with nature. while "ext" is not the short or abbreviated form for any popular thing i am aware of, it could conceivably be an acronym for something which is or becomes popular. not so with .external. > Does it matter to you that "external" has no unambiguous connotation > in lots of languages? If not, then why is the difference between > "alt" and "external" important? If so, why don't we need to reserve > the, ahem, "equivalent" of "external" in every language, forever? since apple has been selling internationally for decades, and their .local has never been translated into non-english character sets, i think it's possible that all internet users and application, for all time, will have the ability to read and type simple english words. -- Paul Vixie _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
