>any domain names you want, but the safe choice for avoiding operational and >policy collisions with >DNS protocol and namespace is to root your chosen domain name space under >.alt"? > >Which technical issues would persist?
I can think of two, plus one non-technical one. The first technical one is whether there is a registry for .alt names and if so how it'd operate. From prior discussion, some people think it should be FCFS to prevent ambiguity. I think that would be utterly counterproductive since it'd lead to a squatter rush, and anyway there's no way to enforce that people use names they've registered. FCFS with unlimited name collisions would be OK, to help people figure out where to get the code to implement various name things they've come across. The other, which is a can of worms, is whether we want to define protocol switches. At this point, there's an implicit switch used for .local at the DNS level, where you give it a name and it returns an IP address that might have come from an A or AAAA record or might have come from somewhere else. And there's another higher level one at the socket level used for .onion, where you give it a name and it gives you back an open socket that might be a TCP or UDP connection to an addresss found in an A or AAAA record, or it might be something else. If there's to be any hope that people could use more than one of these hacks at a time, a protocol switch would be a big help. Finally, no matter what we do, at some point someone will come by with .GARLIC which is like .ONION but stronger and they will say (with some justification) that it's used by a zillion people around the world. "You should have used GARLIC.ALT." "Yeah, I guess so, but we didn't, sorry." Then we'll have to deal with it one way or the other. I hope that .alt will push that day off farther into the future but it's unlikely to push it to infinity. R's, John _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
