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

[Shawn Steele]

> Sure. Part of the mapping definitely happens when characters are
> input: I have yet to meet a keyboard driver that outputs scan codes
> that look like any character set (let alone Unicode) that I know. And
> It certainly all happens before a string of characters is handed off
> to the wire.

This section seems a bit of overkill.  I can talk to my phone and end up making a URL out of that.  Certainly no scan codes are involved :)  I think that the mapping document can start at "there's a Unicode string that needs mapped for IDN.'

> Well, it would not be complete havoc for an application that was
> implementing IDNA to put that lower-case mapping at some point after
> the keyboard and before punycode. In fact, we require it.

Hmm, in DNS, a punycode string like XN--UPPERCASE would still resolve...

> I personally think there's no need in IDNA to do it before displaying
> it on the screen.

I think there's too much emphasis on input & display.  The inputs could be spoken and outputs could be read aloud.  They could be links embedded or cut & paste from existing text.  The reader could be a web 'bot and not human....

It's nice to get started with something for a draft though :)

I also agree with Mark that mappings are a "MUST", not a "SHOULD", or else different browsers or whatnot will end up in different places.

- Shawn