... > >The problem here is that the COMBINING GRAPHEME JOINER only affects > >*enclosing* combining marks (and combining marks *following* an > >enclosing one). > > I do not know what you mean by this. The CGJ makes what it is joining > into a single entity. So if you add a diacritic to it, it should > centre itself above that entity.
See http://www.unicode.org/unicode/reports/tr28/, which says: "On the other hand, where elements are linked by a Grapheme_Link, non-enclosing combining marks only apply to the last base character." That sentence is followed by an example with NUKTA. So Grapheme_links do not affect marks that are neither non-enclosing nor (Brahmic) dependent vowels (the latter is not properly covered by the quoted sentence, but talked about elsewhere in the Unicode 3.0 book). If you will, "non-enclosing" combining marks bind strongly to the last base character, while "enclosing" combining marks (which apparently are Me + Brahmic dependent vowels, though there is no explicit definition) apply to the grapheme cluster also when it is created via a Grapheme_link, e.g. CGJ. (There is also a special case for VIRAMA at the end of a word; but that is a different detail, and I don't know how, or even if, that is now resolved.) /kent k
