Re: [dane] AD review of draft-ietf-dane-openpgpkey-03

[Paul Wouters <[e-mail]>]

On Wed, 25 Jun 2015, John Levine wrote:


> In hashes, you might want to do some back of the envelope calcuations
> about how many CNAMEs you'd need to cover even a modest number of dots
> and case variations.
>
> Speaking of case folding, I note that nobody here seems to understand
> UTF-8 and EAI.

 	"This sort of misrepresentation of what other people [have said] is
  	 extremely unhelpful."

> Really, it matters.  If you think it doesn't, please
> add a sentence saying "This specification MUST NOT be implemented in
> any country where languages other than English and Hawaiian are in
> use" since those are the only ones that can be written in ASCII.

I'm at ICANN53 and talked to some of the people that are well versed
in EAI, such as Asmus Freytag, Hirofumi Hotta and Wil Tan.

There is apparently only one language where lowercasing can change the
meaning of a letter and that is Turkish (with the letter I)

Their recommendation was to only lowercase for ascii and to normalize
everything else, then hash (or base32/split)

Normalization seems to be a proper way of doing things. Some references:

https://tools.ietf.org/html/draft-dainow-eai-email-clients-00#section-5
https://tools.ietf.org/html/draft-ietf-eai-rfc5335bis-08#section-2.2
https://tools.ietf.org/html/draft-klensin-net-utf8-09

So my suggestion is to recommend normalization and refer to
draft-dainow-eai-email-clients and draft-klensin-net-utf8

I still feel that using this ruleset, using a hash seems fine, but if
people really feel that live signing DNSSEC servers talking to live mail
servers is a thing that must be supported instead of that use case being
handed of to a separate webfinger document, I could go with base32/split
as well.

I'll try to run into Warren again here at ICAN and see how and when he
would like me to update the document.

Paul