fully-qualified domain names containing differently-encoded labels ?

[Andrew Sullivan]

On Mon, Oct 25, 2010 at 02:51:38PM -0700, =JeffH wrote:

> Understood. However, in various protocols, e.g. HTTP, one has to handle 
> (e.g. compare) strings matching RFC3986's "reg-name" production, and thus 
> must decompose said strings into labels, do the IDNA magic as necessary, 
> and then re-construct as appropriate and compare (for example).

My strongest advice, then, is to devolve everything into what IDNA2008
calls LDH-labels (i.e. either A-labels, fake-A-labels, or
NR-LDH-labels) and compare those.  That will be reliable, and you
don't have to care whether the putative A-label or U-label actually
qualifies under IDNA2008 rules.

> wrt the latter, fwiw, note that RFC5890 uses the terms "label type" and 
> "label form" essentially synonymously; and the former is explicitly used 
> to denote label types..
>
>    For IDNA-aware systems, the valid label types are: A-labels, U-labels,
>    and NR-LDH labels.

Ah.  I thought you were trying to include other "types", such as
labels that are protocol-allowed under IDNA2003 and not under
IDNA2008.  That was the thing I didn't want to get into: once you're
outside IDNA2008 (which 5890 is talking about), "type" is not useful.
You seem to want to deal with input that _might_ be outside IDNA2008,
so you can't datatype the inbound data into only the IDNA2008 types.
Right?

A

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