Unconditional punycode conversion

[Andrew Sullivan]

On Wed, Mar 09, 2011 at 09:21:12PM +0100, Simon Josefsson wrote:

> This was not clear to me.  Section 5.1 and 5.2 gives the distinct
> impression that section 5 is about the _entire_ domain name process, for
> all kind of labels, from reading the label from the user to looking them
> up in the DNS.

Yeah, reading it especially in isolation, I can see how it comes
across that way.  It can't be doing that, however, because it is
explicitly not updating RFCs 1034 and 1035.  Indeed, if we were
willing to update 1034 and 1035 in order to make UTF-8 labels easier
to cope with, the set of trade-offs inherent would be totally
different.  When you start with the premise, "It's ok if every
computer in the world has to be updated to support this," then a lot
of things become possible :)

> Programs typically doesn't have separate input fields
> for LDH-labels and U-labels, so normal programs needs to be prepared for
> both.  

This is true, but it does not entail that the programs have to treat
every name as the same.  For instance, programs often don't have
different interfaces for entering DNS names and netbios names, and yet
they manage to deal with this.

The idea underlying IDNA2008 is that your IDNA2008-aware application
knows something about A-labels and U-labels and NR-LDH-labels and all
that, and therefore will be able to differentiate them and handle them
differently.  

> Btw, technically what do you mean by a "IDNA2008 label"?  I can't find
> any definition of the term.

This was just a term I made up on the fly for "label types that
IDNA2008 has something to say about", to distinguish it from
"everything else that could go in the DNS".  

For instance, leaving aside the more fanciful UTF-8 labels and so on
(though we know they're in use in some places), consider the
underscore label, like _tcp, used with SRV and other such RRTYPEs.
Those are simply not discussed in any way by IDNA2008.  They are, by
definition, never a possible A-label.  They're also not an LDH label:
they don't conform to the traditional hostname restrictions.  But
they're found all over the place.

A

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