Re: UTS46 "transitional period"

From: Chris Weber <>
Date: Tue, 28 Jun 2011 19:41:22 -0700
I was trying to understand the implementation differences in some browsers and registrars.  Using your example from UTS46

Opera - error, doesn't resolve
Internet Explorer - error, doesn't resolve
Chrome, FF, and Safari - all resolve with no error

The registrar has this domain name registered and returns its own error page.  I tried registering at several registrars and received the error that the domain name was invalid, not that it was taken, but that there was an error, or it was illegal punycode.

And why would Chrome, FF, and Safari resolve this - wouldn't the nameprep process under IDNA2003 have it mapped back during the ToAscii or ToUnicode process?


On 6/28/2011 5:18 PM, Mark Davis ☕ wrote:
It depends on what you mean.

Registrars could and did allow users to enter in labels for registration that would end up getting remapped to other labels before registration. That is, they could let people enter in faß.de or Ö, even though what was actually registered as far as the DNS is concerned was and (=ö, resp.

What was unfortunate is that apparently not all registrars made it clear to the registrants exactly what was being registered.

They should never have allowed (I don't know whether any did or not) registering punycode that was illegal, like xn--a.

— Il meglio è l’inimico del bene —

On Tue, Jun 28, 2011 at 14:40, Chris Weber <> wrote:
On a tangent, under IDNA2003 rules, shouldn't registration and resolution of the punycode form of "faß.de", which is "" be illegal since the nameprep process would mapp the eszett to "ss"?

On 6/27/2011 4:14 PM, Mark Davis ☕ wrote:
Once the registries (of that support the 4 special characters) have adopted bundling/blocking policies, then it will be possible for client software to safely move off of the transitional approach. So it is really dependent on the registries' behavior. Different clients may have different thresholds for when to move, depending on their estimation of when enough important registries have  adopted bundling/blocking policies.

 For example, when all Registries *and* all client software have moved to exclusively support IDNA2008

"All client software" doesn't work, unless you have a coordinate switch date, which people concluded was impractical.

— Il meglio è l’inimico del bene —

Received on Tue Jun 28 2011 - 21:46:14 CDT

This archive was generated by hypermail 2.2.0 : Tue Jun 28 2011 - 21:46:16 CDT