2014-07-02 20:55 GMT+02:00 Jukka K. Korpela <[email protected]>: > I think the idea of using CGJ is more wrong than the idea of using ZWNJ. >
I think exactly the opposite. CGJ brings the distinction that it prohibits the cannical combination. As the resulting string is not anonically equivalent, it is also semantically distinctive. Note that we are in the case were CGJ would be used here just after a base letter. We are not in the situation where CGJ is used between TWO combining characters, where the first one has a *higher* combining class than the second one with a non-zero combining class (this case is where CGJ is used to prevent reordering of combining diacritics during normalization: this case occurs when these diacritics wbich usually don't interact with most base characters may collide on the same position and need an explicit difference, for example when the cedilla occurs above a letter rather than below it and interacts with another diacritic above that letter, and the relative order of the cedilla and that diacritic matters). This is used notably in the Hebrew script (due to the "strange" historic assignment of distinct non-zero combiing classes to most of its diacritics even when they can interact and relative ordering is significant both semantically and graphically). We are also not in the situation where CGJ occurs between TWO combining characters having the *same* combining class, in order to stack them differently. In that case, no reordering occurs, even without CGJ, and the relative order is significant, but there's a distinction between vertical and horizontal stacking. CGJ is also used in cases where there's an enclosing diacritic and another one: should that diacritic be inside or outside the enclosing diacritic? (we have examples in mathematical notations with diacritics liek arrows) The case of CGJ used immediately after a base letter (or after a combining character with combining class 0) encodes either a variant of the base letter, or of the diacritic after CGJ. The only case where CGJ is currently not used is at end of streams ; or just before a base letter (but there could be applications for tricky cases in Indic scripts) where ZWJ and ZWNJ are prefered (to control ligatures and contextual letter forms such as subjoined letters).
_______________________________________________ Unicode mailing list [email protected] http://unicode.org/mailman/listinfo/unicode
