Unconditional punycode conversion

[John C Klensin]

--On Wednesday, March 09, 2011 13:16 -0500 Andrew Sullivan
<ajs at shinkuro.com> wrote:

> On Wed, Mar 09, 2011 at 07:08:45PM +0100, Simon Josefsson
> wrote:  
>> I don't see any of this reflected in RFC 5891.  As far as I
>> can tell, "ab--cd" is permitted since there is no rule to
>> forbid it.
> 
> RFC 5891 isn't the only normative document in the series.  RFC
> 5890 is also normative.  You need to read all the documents
> together.
> 
> But anyway, "Forbidden to whom?"  ab--cd is a perfectly legal
> DNS label.  So, for that matter, is ƒoregoneconclusion and
> 3ßçç~µ.  Any of these could appear in a legal DNS zone
> today.

And this is ultimately why the IDNA2008 specs cannot come out
and say "strings that have '-' in positions three and four are
forbidden always and forever".  One can have a perfectly
reasonable program that interacts with the DNS and that treats
any string of octets as a valid label; it just doesn't conform
to IDNA.  If a program does conform to IDNA, then there are, as
Andrew points out, only three valid types/formats of labels.

>...

   john