On Jan 7, 2015, at 11:12 AM, Andrew Sullivan <a...@anvilwalrusden.com> wrote: > No, I get that, and I understand why we're using the name _instead_ as > an indicator. This is really the same reason people used underscores > with TXT RRs as a selector instead of a different RRTYPE in lots of > cases: you use the interface you have. My only real point is that the > Tor case and, say, the GNS case are really basically different, > because one of them would actually be susceptible to a different URI > scheme whereas others don't work that way. I'm sorry I mentioned > classes; it's a distraction.
Okay, that makes more sense. I think there are really two questions here: 1. Do we need to do something to accommodate existing practice? 2. Ought we (or really they, since we don't have change control over ToR) to change that practice going forward? I think these are actually separate issues that ought to be addressed separately, although they are of course related. You seem to be speaking to the second point, not the first. On the second point, the tradeoffs to consider include cleanliness of architecture, but also usability. I think the best argument in favor of using a different URI schema is actually simply that we want things that don't actually understand how to do onion routing to fail rather than try, since trying using native DNS and HTTP exposes the user to attack on hosts that don't have resolver support for .onion. _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
