IDNA considerations

[chrysn]

I'm reading IDNA (rfc5890 and -bis) right now through an ART-ART assignment … wow, that's headaches.

We're currently describing labels to be turned into percent encoded URIs, so that a CRI that contains "übergrößenträger", "example", "com" as a host name would be converted to http://%C3%BCbergr%C3%B6%C3%9Fentr%C3%A4ger.example.com/.

IDNA post-dates RFC3986 (even 3987) but still talks of URIs as not taking non-ASCII labels.

I'd hate to see any punycode stuff creeping up on us here, so I hope that either

• we find a reference that uses URIs in such a way so it's all fine, or
• we find a place to say that our host components permit non-ASCII labels and are thus IDNA-aware for the purpose of IDNA. (The latter clause probably matters because not all uses of the URI hostname field are necessarily DNS backed, even if there are dots in it, and justifies passing them around without any larger concern for IDNA when not resolving).

[chrysn]

Next step here is to look at https://datatracker.ietf.org/doc/draft-klensin-idna-rfc5891bis/ and related documents whether there's anything to support our notion of "it's really a different layer and doesn't concern us". (Inconveniently, 3987 explicitly mentions IDNA, and not just saying that it's DNS's business how they encode their labels).