Re-opening discussion about Mapping

Vint Cerf vint at
Mon Feb 8 21:15:52 CET 2010


the present mapping document is non-normative; consequently, it is not  
clear who is committing to its implementation. In fact, the text gives  
considerable latitude as to implementation and mapping choices given  
that it is expected that some amount of mapping will be purely user  
interface tactics. A key question is whether and what kind of mapping  
occurs on lookup. We are agreed that no mapping occurs on  
registration. Per my long message, a big question is what constitutes  
a valid label. The IDNA2008 consensus is that only dual A-Label/U- 
label elements should be considered valid domain names (and we note  
that registries can restrict which of the IDNA2008 PVALID characters  
are permitted in labels registered in the relevant zone.

You are asking a question that may not be answerable. There are so  
many variations of user inputs (typing, cut/paste, clicking on URLs,  
marking of strings as domain names - whether they are or not, by  
various application programs, etc. ) and so many potential language/ 
script combination. I believe that many of the choices for local  
manipulation of strings will be invisible to the DNS. What concerns  
some of us is the question of what strings should be passed around as  
uniformly valid, (canonical?) domain names.

At one end of the spectrum is the belief that the only domain names  
that should be used for distribution outside of local contexts should  
be in A-label/U-label form (ie. no mapping should be applied or  
needed). I don't know whether it is possible to achieve that goal  
because people put up documents, create HTML, and do other things that  
cause non-canonical forms to be sent around in the Internet. You are  
familiar with the notion of Fully-Qualified-Domain-Name I know. This  
concept was adopted to avoid the problem of a local "domain name"  
escaping and ending up with the wrong suffix. It seems as if we are  
looking for a kind of canonical formulation for what I will call here  
"PVALID" domain names.  I think the present suite of IDNA2008  
documents says what that is. The mapping document is addressing  
predecessors to that form but in an unspecific context.


On Feb 8, 2010, at 2:33 PM, Lisa Dusseault wrote:

> In order to be able to handle the mappings document, should the WG  
> choose to put it forward, I would like to understand as part of this  
> discussion:
>  - implementation intentions  -- who plans to implement mappings and  
> in what context (e.g user typing? user click on URL?)
>  - clarity -- is it clear which characters are suggested to be  
> mapped, e.g. so that we understand the difference from what  
> characters were required to be mapped in IDNA2003
> thanks,
> Lisa
> On Sun, Feb 7, 2010 at 2:35 PM, Vint Cerf <vint at> wrote:
> Please see the attached, rather longish PDF. It is intended to try  
> to move the discussion forward on the matter of the mappings question.
> The WG has previously agreed that:
> 1. there will be no mapping in the registration process. labels in  
> registered domain names must be in U-label or A-label form,  
> satisfying all the validity criteria of the now-adopted normative  
> IDNA2008 documents.
> 2. The mapping discussion emerging from IDNABIS should not be  
> normative
> Vint
> _______________________________________________
> Idna-update mailing list
> Idna-update at
> _______________________________________________
> Idna-update mailing list
> Idna-update at

-------------- next part --------------
An HTML attachment was scrubbed...

More information about the Idna-update mailing list