I-D Action:draft-ietf-idnabis-mappings-00.txt

Pete Resnick presnick at qualcomm.com
Thu May 28 04:31:43 CEST 2009


On 5/27/09 at 2:05 PM -0700, Mark Davis wrote:

>On Wed, May 27, 2009 at 12:57, Paul Hoffman <phoffman at imc.org> wrote:
>
>>I'm glad that we have a document to work from, but this document 
>>seems upside-down.

As I've said to Paul offline, I both agree and disagree. In the end, 
the document might very well be turned right-side up. But I think the 
discussion about the principles needs to be agreed to (at least in 
part), even if it becomes an appendix or an introduction.

>>Similarly, putting the actual algorithm as an appendix just seems wrong.

The algorithm will be in section 3, not in an appendix. A particular 
backward-compatible mapping belongs in the appendix.

>>On a technical level, the algorithm is wrong in that there are 
>>characters that are legal in IDNAbis that will not make it through 
>>the algorithm. We *need* to come to agreement on the mapping of 
>>those, and we need to do so sooner rather than later.

Absolutely agreed.

>    These are the minimal mappings that an application SHOULD do.
>
>The user of SHOULD rather than must means that, for example, one 
>still wouldn't know what a IDNAbis-conformant browser would do with 
>href="http://Übersetzung.de". Would it interpret it as 
>übersetzung.de or just fail?

No, that is absolutely false, and a misreading of RFC 2119. If you 
SHOULD do the mapping, that means "that there may exist valid reasons 
in particular circumstances to ignore a particular item, but the full 
implications must be understood and carefully weighed before choosing 
a different course." In the case of mapping "Ü" to "ü", I can't think 
of a valid reason to choose a different algorithm. But I can 
certainly think of a valid reason to choose to deal with "ß" 
differently in different contexts, and I can think of others.

>By having no firm requirements...

Nonsense. There are absolutely firm requirements. "You SHOULD 
implement this algorithm" means "you MUST implement this algorithm 
unless you have a darn good reason not to." They are not absolute 
requirements, if that's what you mean.

>...it leaves the issue open, and we are no better off than if this 
>document weren't there.

If your criteria for the usefulness of this document hinges on there 
being one and only one mapping from keyboards to domain names, the 
two of us will simply have to agree to disagree. I can't speak for 
others in the WG.

pr
-- 
Pete Resnick <http://www.qualcomm.com/~presnick/>
Qualcomm Incorporated - Direct phone: (858)651-4478, Fax: (858)651-1102


More information about the Idna-update mailing list