On Feb 22, 2015, at 8:49 AM, Andrew Sullivan <[email protected]> wrote: > > On Sun, Feb 22, 2015 at 08:31:08AM -0800, Paul Hoffman wrote: >> >> As I said earlier, I think that "labels in domain names should be encoded in >> Punycode" would be sufficient, and would eliminate your problem with >> _labels, I believe. >> > > What exactly makes you decide that underscore labels aren't domain > names?
A mistake in me remembering which label the "xn--" prefix was applied. :-( (That's really a :-(:-( given that I am the co-author on that document...) So, n'r mind what I proposed above. A better rule would be "express labels in the wire format, except where the wire format is not printable ASCII, in which case use \Dxxx escaping". That is, IDNA would have nothing to do with this. As for Måns original question: converting wire-format IDNA to some encoding of Unicode characters is unstable because some registries use IDNA2003 rules, others use IDNA2008 rules, and some labels in the former can't be represented in the latter. Yes, IDNA2008 obsoleted IDNA2003; that is irrelevant to the a good chunk of the operator community. --Paul Hoffman _______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
