RE: [idn] Reality Check

[Martin Duerst <duerst@w3.org>]

At 18:25 01/07/14 -0700, John C Klensin wrote:

>It seems to me that your argument, with a very slight recasting
>or none at all, would make at least as strong as case for one of
>the major, "solve more of the problem" approaches -- search
>systems

Yes indeed, except that we know quite well how to build and
run the DNS (including where we have to be careful,...),
whereas we don't really have any experience at all of how
to build an Internet-wide reference search system. It may
be that somebody comes up with some great idea tomorrow,
and such a system will fully exist in a year, but it also
may be that we never get there.

Another point is that search systems don't provide the same
as the DNS, namely lookup. There is no guarantee that you
get to the right place (or nowhere) directly. So 'solving more
of the problem' may cut off some of the problem space.


>and new DNS Class included-- than for embedding UTF-8 in
>the existing, Class=IN DNS.

A new class is one of the many (unfortunately too many) options
to upgrade DNS to UTF-8, on one end of the spectrum, whereas
'just send UTF-8' is on the other end.



>To pick up on one part of your example, the companies that
>managed to upgrade their mail systems to provide Internet gateway
>capability did not put any of their existing uses at significant
>risk,

Using UTF-8 is a clear add-on, isn't it? Where is the significant
risk? With 'just send UTF-8', there is the risk that a DNS server
blows up, or returns something else e.g. because it folds the most
significant bit in each byte to 0. But both of these behaviors
are clearly against the DNS specs. If a server blows up, it was
a risk 10 years ago already (if it existed then).
And 'any of their existing uses' is definitely NOT put at risk.
NONE of the UTF-8 proposals is putting at risk the resolution
of ASCII-only domain names.


Just a few observations,      Martin.