Re: [idn] Transparent vs. ACE representations (was We are quibbling about WHAT?)

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

David Hopwood <david.hopwood@zetnet.co.uk> wrote:

> The thesis behind the ACE-only proposals

What ACE-only proposals?  I think the strongest ACE supporters
merely say that you have to obey existing standards.  If a protocol
specification (like the URI spec, or the SMTP spec, or the TLS spec)
says that the domain name in a particular slot may contain only certain
characters, then ACE is the only way you can use IDNs while still
obeying the standard.

But new standards can of course specify looser rules.  I don't think
anyone here wants to prohibit new protocols/interfaces from using UTF-8
(or their favorite charset).

> most things don't assume that the representation is ASCII

As Dan Bernstein's recent tests have shown, lots of software stops
working when domain names contain bytes greater than 127, including
browsers, resolvers, mail user agents, and mail transfer agents.

And one thing's for sure:  No deployed software today knows how to
properly compare two IDNs (which requires nameprep).

Even if you're right about most things not breaking on 8-bit domain
names, things are connected to each other.  Domain names get passed
around.  If you feed an 8-bit domain name to an IDN-unaware agent, it
might not break that agent, but it's probably going to break something
else a little later.  Therefore 8-bit domain names ought to be passed
around only among IDN-aware agents/protocols, and converted to ACE when
interfacing with IDN-unaware agents/protocols.

AMC