Re: [idn] Debunking the ACE myth

[Martin Duerst <duerst@w3.org>]

At 15:54 01/07/19 +0200, Patrik F$BgM(Btstr$B‹N(B wrote:
>--On 01-07-19 16.47 +0900 Martin Duerst <duerst@w3.org> wrote:

>Reason for this is that ACE is backward compatible with the standards we
>already have, such as RFC 2821 and RFC 2822. If we have a solution which
>points at UTF-8 as the transport, then the RFC's specifying the application
>layer protocols have to be updated -- and proper handshake (if needed) and
>downgrade to ascii encoding (if needed) is defined.

Yes. But that's each protocol's business, not the business of the DNS,
as you say below.


>I.e. the problem with using non-ascii in the labels in DNS, is that it is
>really easy to get what I would call leakage into the application layer
>protocols which can not handle other characters than ASCII.

As Dan and others have shown, that's really easy also for ACE.
Even if we get all software doing it as planned, users will leak.


>How to handle Unicode and how to reach the goal of using UTF-8 encoding in
>the future is something I rather see each wg which have knowledge about the
>specific protocol. Regardless of how easy or hard it is.

Will they ever do anything? NNTP has done most of their job (headers all
in UTF-8), others may for one reason or another never get to it. Should
we punish those that do their job by having to do more, or should we
provide motivation for those that want to do the job?

Regards,   Martin.