Two-step mapping

[Andrew Sullivan]

On Sun, Jul 26, 2009 at 03:21:40PM +1000, Wil Tan wrote:

> user given specific user requirements, and *avoids the two-step mapping* of
> the original protocol.
> 
> 
> In practice, if the application were to do any mapping, it is likely to
> perform a two-step mapping anyway:
> 1. As earlier paragraph suggests, the first step of getting the input as
> Unicode code points is likely the function of the input method or operating
> system event loop.
> 2. The application then performs its local mappings.

John would be the authoritative source of what the intent behind that
text is, but my reading of the "two steps" in IDNA2003 is more like
this:

1.  Do whatever is needed locally to get things into sensible shape.
(This would include 1 & 2 above, I think).

2.  Map inside the protocol.

I think the complaint is that there is truly-local mapping, and then
there's protocol-level mapping in IDNA2003, and the second step (in
the protocol) is what is desired to go away.  (That said, I agree that
the main focus needs to be on the local sensitivity IDNA2008 offers,
because different mapping rules in different contexts are allowed.)

A



-- 
Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.