Re: [idn] Debunking the ACE myth

["Eric A. Hall" <ehall@ehsco.com>]

> > The dual-mode model helps you deliver on whatever you decide you
> > can|wan't to do. It won't break anything on it's own.
> 
> it will break things merely because it requires additional software
> changes. changes invariably introduce new bugs.

> poorly-implemented applications are a given, I'm afraid.
> and the more complex we require them to be, the more likely they are to
> be buggy.

These are all arguments against ACE transliteration (and IPv6 and just
about everything else for that matter).

The only way that UTF-8 IDNs will get into the legacy path is if a
particular implementation goes out of its way to use them where it
shouldn't. It will have to explicitly call the new API and then explicitly
pass data that is not supported by the application protocol. This scenario
is completely unrealistic.

You are essentially arguing that if an application makes two unrealistic
errors then UTF-8 IDNs will break the application.

Your initial comment was:

 | in most cases, it's simply not feasible to separate the use of IDNs
 | from the use of ASCII DNS names and negotiate or convert at the
 | boundaries.

You have yet to prove this point. On the contrary, I have shown that it
requires explicit effort for problems to occur, and even then the problems
are not due to anything inherent with the protocol itself.

Please, let's drop this FUD and move on.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/