The quotes below are from Bruce's remarks in the real-time captioning of the GNSO Public forum. I am not being at all critical of the way anything was expressed in that context, but as we proceed with the development of formal IDN policies we will need to become very precise in the terminology we use.
at the moment, the IDN guidelines suggest using a single script to the left of the dot, so you don't have a mixture of, say, English or Latin scripts and Cyrillic scripts in the same name or mixtures of different scripts.
The guidelines say not to mix scripts in a single _label_, unless the language being represented is normally written using multiple scripts. This allows for situations where there may be arguable need for having multiple scripts in a domain name even when the intended language is normally written using only one of them. (Examples of such cases have been provided by representatives of the Unicode Consortium, if memory serves, primarily via their own e-mail discussion list.) This consideration is entirely separate from the current restriction to the exclusive use of Latin script in top-level labels.
if I have English script, then that connects to an English script string in the root. And then if you wish to use Chinese script, then that's associated with Chinese script TLD in the root. Or do you allow mixes? And we have today, you can have English or Chinese -- or you certainly can have today a Chinese script dot English script.
The English _language_ is written using the Latin _script_. A large number of additional languages are written using that script. Some of these languages are more reliant than others on the ability to indicate diacritical marks in addition to the shared repertoire of basic characters (with it being a common misconception that English is entirely free from such need), but ASCII can comfortably be used for quite a few of those languages. Although we do urgently need to get beyond the constraints that it imposes, we should keep in mind that ASCII is not a synonym for English. Many of the individual languages that share the Latin script are used in more than one country. The same can apply to the different languages that are written -- listing only a few of many possible examples -- with the Arabic script (which, unlike the term English, does designate both a language and a script; Latin script is frequently called Roman script precisely to avoid this type of ambiguity), the Chinese script, the Cyrillic script, the Devanagari script, and all the other scripts that either are already under consideration or are going to appear in short order as we proceed with the internationalization of the namespace. There is certain to be some discussion of the extent to which countries may "own" the scripts used within their borders, but there are astonishingly few cases where there is any unique relationship between script and country. Corresponding conditions apply to languages. The first version of the IDN Guidelines was marred by its having been written under the assumption that the terms "language" and "script" were synonymous, with the preferential use of the former to avoid confusion (despite clearly differing opinions about the likely consequences of doing so). The 2.0 revision was prepared specifically to rectify the difficulties to which this led. I hope that I am not alone in feeling that the terminological stringency now reflected in the Guidelines should also be conveyed by the material that Council prepares. /Cary