draft-klensin-idnabis-protocol-04 section 4.5

Paul Hoffman phoffman at imc.org
Thu Mar 27 17:36:45 CET 2008

At 1:17 PM +0100 3/27/08, Harald Tveit Alvestrand wrote:
>--On Thursday, March 27, 2008 10:55:04 +0100 Simon Josefsson 
><simon at josefsson.org> wrote:
>>This section reads:
>>    The resulting U-label is converted to an A-label (i.e., the encoding
>>    of that label according to the Punycode algorithm [RFC3492] with the
>>    prefix included, i.e., the "xn--..." form).
>>That assumes that no U-label will be translated into a LDH-label.
>>In IDNA2003 some U-labels are translated to LDH-labels, for example:
>>ToASCII(josefßon) = josefsson
>>ToASCII(dªtªkonsult) = datakonsult
>>Note absence of xn-- prefix and punycode data.
>>Is the intention that these strings will not map to the same LDH label
>>in IDNA200x?
>As long as we keep the "no mapping" principle, the intention is that 
>these strings will be rejected by IDNA200x.

I fully disagree with Harald, and this is fundamental to the 
definition of the protocol. Those strings would be rejected if the 
application did not do sufficient mapping before doing a lookup as 
specified in Section 5.3. An implementation that wants to 
interoperate with IDNA2003 for labels that were mapped in IDNA2003 
would do some mapping.

Thus, "these strings" do not need to be "rejected" by IDNA200X; they 
would only be rejected if the user interface in Section 5.3 was 
insufficient to match user expectations. This is a good reason to 
consider the normative language for Section 5.3 to be SHOULD, not MAY.

More information about the Idna-update mailing list