<div>John,</div><div>I am glad we at last discuss semiotics in a proper way. You asked if we should enter typography issues. The problem we have from the very beginning is that internationalization constrains the human intellectual semantics to the Anglo-saxon pragmatics, added with some Internet usage/standardization typographical syntactics (BlueBell, is actually using a virtual space). </div>
<div><br></div>Would there be a problem in creating another type of characters as "IGNORED", meaning that they can be typed but will be ignored by the DNS?<div>Also, could we not use DISALLOWED code points as a temporary pool of TEMPORARY code points for typographies not supported (yet) by Unicode?</div>
<div><br></div><div>jfc</div><div><br></div><div>PS. As far as I understand punycode supports upper/lowercases. Therefore French and other languages have much less problems in a non-case folding, with IGNORED and TEMPORARY classes. What I observe is that contextual rules are trying to replicate the contextual usage practice and may confuse this practice, actually diminishing security. They also reduce flexibility: I am not sure that the semantic/security trade-off is positive. I know this would change the general perspective of IDNA and also open new possibilities for attacks. This is why I think the best is to use a regular "xn--" IDNA and an extended "xx--" IDNA, with possible "x.--" variations when needed. The use of the same punycode algorithm should greatly increase the security. </div>