Some bullets for the IDN Guidelines (draft)


Mats Dufberg
2015-11-17


What should the Guidelines contains? Who is the audience? Below I have take an broad perspective. The bullets below are meant to be useful for any TLD that have or plan to have IDN support. I could also be useful for second level domains. Each bullet should be filled with explanations, justification and reference. They are not ordered or grouped below, just listed.

This is just a start.


* Always base the IDN table on one or several linguistic groups and meet their needs.

* Never include a code point unless it has IDN property Pvalid, ContextJ or ContextO.

* Implement the contextual rules that code points with IDN property ContextJ and ContextO require.

* Implement appropriate context rules for code points with Unicode category Mark.

* Investigate the need for contextual rules for code points with Unicode category Modifier Letter.

* Investigate the need for contextual rules for code point of script property value Common or Inherited.

* Implement variant management if labels with different Unicode scripts with homo-glyphs, e.g. Greek and Cyrillic, are permitted.

* Implement contextual rules for HYPHEN (U+002D) in the U-label.

* Implement appropriate contextual rules for RTL labels.

* Never mix LTR and RTL code points in the same label.

* Investigate the need for variant management of European digits, Arabic-Indic digits and Extended Arabic-Indic digits, if more than one category is permitted.

* The general rule is not to permit mixing of code points of different explicit script property value in the same label.

* Mixing of points of different explicit script property value in the same label should be justified on cultural or linguistic usage.

* Code points of script property value Common or Inherited should only be used if it can be justified.

* Be conservative when constructing the IDN table, rather exclude code points than risking homo-glyph attacks.

* Be conservative when adding new IDN tables, it is easier to add than to remove.

* Publish the IDN table in the LGR format.