Re: [idn] Who supports UDNS (ACE+UTF-8)

[Marc Blanchet <Marc.Blanchet@viagenie.qc.ca>]

At/À 06:34 2001-06-30 +0000, Adam M. Costello you wrote/vous écriviez:
>"Eric A. Hall" <ehall@ehsco.com> wrote:
>
> > I think that it is probably also feasible to adopt a "continuation"
> > syntax in ACE as part of a backwards compatibility.
>
>This has been proposed before.  It's certainly a clever idea, but
>it's not clear that it wouldn't create more compatibility problems.
>Applications sometimes pay attention to the number of components in a
>domain name (the web cookie protocol is one example), and sometimes
>chop them up.  If the name appeared to have a different number of
>components when viewed by an IDN-aware application versus an IDN-unaware
>application, there could be subtle interoperability problems.
>
>Consider an IDN FOOBAR.org whose ACE is ace1.ace2.org.  An application
>might be faced with the question of whether FOOBAR.org is a subdomain of
>ace2.org.  Is there a right answer?
>
>The nicest thing about the ACE approach is that it obviously breaks no
>software or protocols.  That would no longer be obvious if continuations
>are possible.  My get feeling is that it would be more trouble than
>it's worth, but if you want to pursue this, please try to work out some
>semantics that are consistent and demonstrably safe.

I concur with Adam. Continuation of idn over multiple labels looks clever 
but makes a more fondamental shift in the dns where labels are no more 
separated by a dot.  To me, it looks really dangerous.

Either we stick on ace with restricted-length encoded idn (i.e. keep the 
limit of 63 octets), or we use edns or a new class to signal this new 
"feature". (Even with that, on the application side, still looks crazy to me).

Marc.


>AMC