Re: [idn] proposals and deadlines

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

Edmon <edmon@neteka.com> wrote:

> These in turn (especially ACE leakage) would encourage the
> application/resolver developer to eventually implement EDNS.

You'll have to walk me through the argument, because I just don't get
it.  DNS servers will be motivated to answer ACE queries correctly
because a lot of queries will be in ACE format, at least at the
beginning.  It doesn't matter whether they perform ACE conversion
on-the-fly, or at zone-loading time, or if they rely on administrators
to use preprocessing tools on the zone files, DNS servers will be
motivated to handle ACE queries somehow or other.

Therefore, resolvers and/or clients will have a choice between
attempting to use a DNS extension and falling back to ACE when it fails,
or just always using ACE.  By the previous argument, the latter option
will work anyway, and is easier, so it will probably be used in many
cases.  And that in turn will keep alive the motivation for DNS servers
to handle ACE queries correctly.

I'm not saying the DNS extension is bad, I'm just explaining why I
wouldn't expect it to become ubiquitous quickly.

"Eric A. Hall" <ehall@ehsco.com> wrote:

> a) the resolver doesn't encode anything, it just passes whatever the
> client application passed to it via one of the APIs

Okay, I thought you were saying that if the EDNS query fails, then the
resolver would perform ACE transformations on behalf of the application.
Now I guess you're saying that the resolver would simply return an
error, and the application would try again using the old resolver API
and the ACE transformations.

This just moves my argument up a layer.  Now the application has a
choice between trying the new API and falling back to ACE, or just
always using ACE with the old API.  The latter is simpler and works just
as well (and is more portable).

If you want to entice me to use the new resolver API, you're going to
have to offer more incentive.  Having the resolver perform the ACE
conversions for me would do it.

AMC