Another field in the DNS

Andrew Sullivan ajs at
Mon Mar 2 20:31:13 CET 2009

On Mon, Mar 02, 2009 at 11:06:34AM -0800, Shawn Steele (???) wrote:

> Erik suggested "another field in the DNS", and I have a similar
> (lack of) knowledge of exactly what is doable, but it seems like we
> need a "display name" for a resolved domain name.

If you need a display name, you need some way to get it.  There are a
few ways to do this:

1.  The in-band suggestion John made of just putting more metadata in
HTML files.  See upthread for the objections.

2.  A special file, fetched via http, that gives the display mapping.
See upthread for objections.

3.  A new RRTYPE that provides a way to pass along this additional
information.  Aside from the objections upthread, remember that some
of Microsoft's APIs don't properly support the Unknown RRTYPE.  I've
heard people claim that it is effectively impossible to provide that
support -- I've never seen the code & probably wouldn't understand it
if I did -- which means that, for this approach to work, we may need
all the Windows machines in the world to be upgraded.  That's even
worse than requiring all the resolvers in the world to be upgraded. 

4.  Invent DNS2 and require that it include formatting information
this way.

I've so far heard nothing in the discussion of this line of thinking
that made me optimistic about quick deployment, but if you have an
idea I'm sure people would be delighted to hear it.
> when displaying the name.  Even ASCII casing can have that problem.
> AAA's links say, however IE displays them as aaa.

Even if you type into the URL bar?  Remember, DNS is
case-preserving and case-insensitive, so matches; but
if you ask for that's what you should get back.  I just did
some experiments to be sure, and when I ask for I get
AAA; aaa then aaa; and even AaA then AaA.  So DNS at least looks to be
working according to its usual rules.  I have no opinion about what
DNS query an application ought to make when it tries to follow a URI

> So if IDN had rules where names mapped to a form that could be
> resolved, but rDNS records could return unmapped forms (so long as
> they mapped back to the normalized form), then some of these display
> type preferences would be fixed.
> Might also encourage people to make their reverse-mapping match too :)

The reverse tree may contain more than one PTR, don't forget.  Also
note that there is apparently less than universal agreement that the
reverse tree is good or even useful (this is a sore point with me,
actually).  It certainly isn't required to be there.


Andrew Sullivan
ajs at
Shinkuro, Inc.

More information about the Idna-update mailing list