Re: [idn] Reality Check

["Adam M. Costello" <amc@cs.berkeley.edu>]

"Brian W. Spolarich" <briansp@walid.com> wrote:

> MTAs (sendmail, etc.) generate MX lookups directly instead of through
> the resolver.
>
> Anything that uses DNS data but doesn't do get{host|node}by{name|addr}
> doesn't use the system resolver.

Okay, thanks.  It sounds like there are indeed some significant
applications that speak DNS directly, but going through the resolver
is by far the common case.  (Hmmm, now that I think about it, I bet
the inability of the resolver to provide access to arbitrary resource
records is the main reason why so few resource records have been
deployed.)

Edmon <edmon@neteka.com> wrote:

> What I meant was the application using the DNS information then
> communicating to a destination server such as an http server.  That
> is, after they obtain an RR from the DNS, how they use the information
> to then connect with the destination.  Should we not allow this to be
> in UTF8 in the future?

I don't follow.  The application gets the domain name from the user or
an href attribute or wherever (not from DNS).  The application hands
the domain name to the resolver, and gets back an IP address.  The
application then opens a TCP connection to that IP address and sends the
HTTP request, which includes the original domain name but not anything
that came back from DNS.

In some other scenarios, applications might indeed receive domain
names from DNS and copy them into application messages.  But the
application is almost always getting the name from the resolver, not
from DNS directly.  Therefore, if you want to make things easier for
the application by giving it UTF-8 domain names, you need a resolver
interface that can return UTF-8 when invited to do so by the application
(but the resolver should return UTF-8 only if explicitly invited by the
application, otherwise it will break existing applications).  Once you
have a UTF-8 resolver interface, it doesn't matter what protocol is
running between the resolver and the DNS server, because the application
can get UTF-8 from the resolver regardless.

> Are you suggesting then that the discussion on using a UTF8 extension
> for IDN in the DNS would be outside the scope of the IDN wg?

Not at all.  I am suggesting that if your goal is to make life easier
for application programmers by allowing them to use UTF-8 internally
and never use ACE, then the way to accomplish that goal is to extend
the resolver interface.  You can also change DNS, but that is less
helpful to applications (it mostly affects resolvers and DNS servers)
and therefore less urgent.

AMC