The Future of IDNA

Andrew Sullivan ajs at
Fri Mar 20 07:11:44 CET 2009

On Thu, Mar 19, 2009 at 10:05:55PM -0700, Erik van der Poel wrote:
> We have heard from the second group, via Ken. And, I'm not convinced
> that we should leave tonos on the letters before registration for a
> small group, that deals with classical text, in the DNS.

But you seem to be missing my point, which is that your proposal
forces us to make a choice about which of the groups get the
capability at all; the current IDNA2008 proposal does not require
that, because it leaves the protocol neutral on the subject, thereby
allowing different zones to work in different ways.

This is consistent with the traditions of zone operation, which have
usually left the decision-making to the zone operator as much as

> Let me point out that the IDNA2008 drafts currently suggest the same
> tactics (bundling and multiple lookup). I don't think the author
> expects those transitions to last until the heat death of the
> universe.

Look, we have root server IP addresses that have been out of service
for _many years_ that still get substantial DNS query traffic.  At the
same time, root operators routinely observe that the "legitimate" root
traffic is statistical noise compared to "garbage" queries that happen
to land at the root servers, because of "safe" decisions made by
various vendors based on their understanding of what whatever BIND
version they happened to have nearby did.  (It used to be even worse
before AS112 got going.)  There is no reason on empirical grounds to
suggest that any strategy involving additional DNS redirections,
additional lookups, &c &c will ever go away completely.  This means
that whatever burden we place on zone operators is never temporary:
it's a permanent part of their lives.  It is true that these
transition strategies will gradually mean fewer and fewer such
additional queries, as people update their systems: the traffic will
diminish.  But it will probably never get to zero (and it will never
be possible to prove a potential of no traffic).  The only reason this
might be worth the cost is because (1) IDNA2003 needs to be replaced
and (2) the IDNA2008 proposal is supposed to avoid having to do this
again.  If either of those premises is plainly false, all of this work
is a waste of time from the point of view of a DNS zone operator.

> No, we only need to deal with actual requests, not hypotheticals.

And what about those who aren't here "actually requesting" -- or here
_right now_ actually requesting?  First, the point of separating
policy from protocol is so that you can "bake in" as little as
possible to the protocol, thereby allowing the policy to change if
need be.  Second, the WG said it wasn't going to go through all of
Unicode, deciding one code point at a time what gets to be in the
club.  If we start to make the list of exceptions both endless and
dependent on who shows up and asks, we will have a deluge of requests
immediately.  I'd bet also we'll get people who didn't hear in time,
and who press for a new round of exceptions within seconds of the
publication of the new, actual-request-granting RFCs.

> How far do you want to take that argument? Some of the French are
> asking for

Even if some don't like it, the Ecole decision was taken many years
ago.  There is just no way to change (or, if you want, "fix") that and
still have anything that vaguely resembles DNS.  As I've suggested
more than once, there are some (I might be one of them) who would not
argue too hard against the proposition, "The DNS is fundamentally
broken, and a replacement ought to be designed and deployed taking
into account the lessons we have learned."  But I suspect that effort
is at least as large as IPv6.  I'm not sure the entire world will wait
that long to fix the IDNA2003 deficiencies.  And even if we were
willing to open a worm-can of those ginormous dimensions, I am pretty
sure that this is not the WG chartered to deal with the resulting red
wrigglers.  If that's what you want to do, you need to get to work on
a BoF proposal for Stockholm.

I get the feeling that we are now revisiting the same arguments
repeatedly.  I therefore think we should try to draw this thread to a


Andrew Sullivan
ajs at
Shinkuro, Inc.

More information about the Idna-update mailing list