Impact of Punycode

Andrew Sullivan ajs at
Fri Mar 26 03:59:38 CET 2010

On Fri, Mar 26, 2010 at 02:01:27AM +0000, Shawn Steele wrote:
> We had UTF-8 labels, that's the same as before. Now we have UTF-8 labels that are equivilent to ASCII lables (punycode).  That's way different.  Before the labels were mostly unique (if you don't count mapping).

Uh, no.  They're _not_ different.  You're conflating layers here.
Please see John's referral to the plenary.  Here's some more (my last)
on this subject:

_In the DNS_, binary labels with UTF-8-shaped octets (or any other
any other 8-bit stuff you want to put in there) are simply 8-bit
labels.  They don't conform to the LDH rule, never did, and never were
even remotely possible hostnames for some things (e.g. mail host

_In the DNS_, IDNA was a way to make something that in the
presentation layer looks like Unicode work (again, in the DNS) as
LDH-conforming labels.  I know that's not news to you.

For Internet-talking things that treated category 1 above as though
the 8-bit labels were "domain names" in LDH-conformant sense of the
meaning, they made a mistake.  The labels didn't actually work that
way, because several applications didn't.  Not all, mind, but some.

IDNA of any kind inserts a layer above the DNS but possibly below the
application interface that converts one kind of input (U-labels) into
another kind for the DNS (A-labels), and conversely.  Now, it might be
true that changes in IDNA, or from binary DNS labels to A-labels,
require changes to that software in ways that are inconvenient.

Moreover, as Pete Resnick said to me this week, part of what IDNA2008
did was move some things from one side of an interface (mappings in
the IDNA protocol) to another side.  But that wasn't an accident: it's
a backward-incompatible change because we'd concluded that the old
approach was harmful.

Yes, it's very inconvenient for our purposes that the DNS has the
protocol, history, and operational practice that it does.  But if the
goal is actually to solve user problems, then this pain is in fact
smaller than "reformat the Internet".


Andrew Sullivan
ajs at
Shinkuro, Inc.

More information about the Idna-update mailing list