The folded form (foldDiacritics: NFD, drop combining diacritics,
lowercase), then one symbol per code point: a-z -> 0-25, 0-9 -> 27,
other ASCII -> 26, everything from U+0080 up -> its own code point.
Section B.2 defines the symbol sequence as continuing from the folded
form, so this must reuse foldDiacritics rather than repeat it -- two
copies of the first three steps could drift, and the index's
never-lose-a-match guarantee (Section B.9) rests on them being
identical.
The folded form (
foldDiacritics: NFD, drop combining diacritics, lowercase), then one symbol per code point: a-z -> 0-25, 0-9 -> 27, other ASCII -> 26, everything from U+0080 up -> its own code point. Section B.2 defines the symbol sequence as continuing from the folded form, so this must reusefoldDiacriticsrather than repeat it -- two copies of the first three steps could drift, and the index's never-lose-a-match guarantee (Section B.9) rests on them being identical.