- 2, Latin transliteration of the native script name: kannaḍa (Also, please use a consistent transliteration scheme in the document.) - 3.4.7, “For 3.4.7.4 there could be cases involving …”: The discussed cases and §3.4.7.4 are not relevant. - 3.4.7, “… hence this is explicitly prohibited by the NBGP”: Not necessary. Just think about writing a note in a limited space then inter-word spaces are extremely narrow — do users have to modify words’ spelling to avoid vowel letters following a consonant with halant? - 5.3: What does “Does not belong to Kannada” even mean? U+0CBC KANNADA SIGN NUKTA is a Kannada grapheme, just not commonly used. U+0CD5 KANNADA LENGTH MARK and U+0CD6 KANNADA AI LENGTH MARK are technically used part of vowel sign character’s canonical decompositions, just not used independently and IDNA2008 requires NFC. - 6.1, “There are no variants within the Kannada script.”: Preconditions are WLE and the limited character set. - 7: A comprehensible pattern for other reviewers to refer to: `C[M][B|X] | V[B|X] | CH` (consonant clusters analyzed as a consonat preceded by one or more `CH` occurences). - 7, Rule 5: Unnecessary restriction. - Appendix II: This is important discussion about required usage of ZWJ and ZWNJ. Should be included in the main text instead of in appendix. Also the rationale/excusing of the lack of ZWJ/ZWNJ is weak and riduculous. - The proposal should discuss the inconsistent encoding and rendering of <ra, virama, ra> /rra/. The preferred rendering form should be glyph sequence <<ra base, ra vattu>>, but in most implementations <ra, virama, ra> yields <<ra base, reph>> thus requires a ZWJ, as in <ra, virama, zwj, ra> (legacy logic) or <ra, zwj, virama, ra> (Unicode recommendation), to trigger the preferred form. Best, 梁海 Liang Hai https://lianghai.github.io