Re: Bundling vs Mapping

[cross-posted to public-iri]

At 07:33 09/02/25, Patrik F$BgM(Btstr$B‹N(B wrote:
>On 24 feb 2009, at 23.18, Mark Davis wrote:
>
>> According to IDNA2008, uppercase Unicode is simply illegal. href="
>> http://B$B—D(Bher.de <http://b%c3%bccher.de/>" would simply fail.
>
>Mark, IDNA2008 is talking specifically about what characters you can  
>use in the dns protocol (as U-labels, and encoded as A-labels). For me  
>it is quite something different than saying whether a certain URI will  
>fail or not.

I agree with what Mark said in his followup message: We need to look
at the whole picture. Being able to claim that we created an IDAN2008
that works in isolation won't help anybody.


>If we talk about domain names, lets talk about domain names. If we are  
>to talk about URIs, lets talk about URIs.
>
>I have always been a supporter of your document that I have not seen  
>any update to (the embryo of an I-D) on how to do mapping and use of  
>IDNs in URIs.

In URIs, IDNs are currently only supported via the fact that for
generic URIs, RFC 3986 specifies that %-encoded bytes in the 
reg-name part are to be interpreted as UTF-8, and must be
converted to "IDNA encoding" (read: punycode) for resolution
via the DNS. I do not expect that RFC 3986 (a Full IETF Standard)
will be updated to say that as part of decoding the %-encoding,
one has to do some kind of mapping.

As for IRIs, I plan to put out a new draft next week. The previous
one expired because of the whole copyright hickup.

What I currently plan to say in that draft is that because we are
not sure how much mapping will be done at what point, everybody
SHOULD produce and pass around IRIs with IDNs in a form that doesn't
need any mapping anymore, and that for the IDN part, some mapping
MAY happen. Given the currently deployed infrastructure, I think
that's the best thing I can do. If anybody has some better
ideas, please feel free to tell me.

You will realize that if we think that mapping for IDNs should depend
on user preferences that may be changing, taking mapping out of IDNs
also means keeping it out of IRIs.


>But URIs are not domain names. To be able to move forward, I think we  
>should separate the various issues here.

We should separate these issues where they need to be separated.
But we should think how these things work together to make sure
it will work overall.

Regards,    Martin.



#-#-#  Martin J. Du"rst, Assoc. Professor, Aoyama Gakuin University
#-#-#  http://www.sw.it.aoyama.ac.jp       mailto:duerst@it.aoyama.ac.jp     

Received on Wednesday, 25 February 2009 05:35:29 UTC