Re: [idn] idn-uri

["Eric A. Hall" <ehall@ehsco.com>]

Bruce Thomson wrote:

> Well, my thought is that we do have to nail down how idns will appear
> in HTTP headers and reponses. Escaped UTF-8 is one way; ACE is another.

One other point, the names which are used in protocol operations (HTTP
headers and responses) can be given specific rules which can be different
from the rules that govern URLs. They can also be converted as required.

 1) users will enter URLs in the preferred charset, standards be damned

 2) URLs with IDNs will have to be converted to a canonical encoding
    whenever they are processed, regardless of whether they are
    provided in this form by software or the user.

 3) The protocol execution of the URL data is subject to the protocol
    rules and not the URL rules.

This may mean that a URL is entered as http://<iso-2022-jp> is converted
to http://<url> by the client for processing is converted to http://<utf8>
for execution. Similarly, a URL is keyed as mailto:<iso-2022-jp> is
converted to mailto:<url> is converted to mailto:<ace>.

I would also put the conversion between (1) and (2) at the viewing system,
and not at the generating system. There is no reasonable expectation that
HTML coded in a localized text editor will do this conversion, so the only
place that conversion can occur will be on the rendering system, when the
URL is parsed. Although it would be beneficial if the generating system
does this work first, and this should be encouraged, it is an unreasonable
expectation to assume that they will.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/