So far the linguistic atlases I have seen extensively use this combining letter mechanism, with diacritics changing the meaning of the combining letter or of the base letter.
There are a whole lot of notations that could simply be base combining letter + combining diacritics, but if you consider their meaning, would have to be encoded a their own combining letter with diacritics on the model of combining a-umlaut, etc... This is not coherent with the decomposition model. Having some of the combining letters with diacritics encoded as single characters and other combining letters with diacritics encoded as combining letters with separate combining diacritics based on what meaning they have is as erroneous as having new precomposed characters because of their meaning. Would we encode ɛ̱ because it is a specific meaning in some language? Obviously no. So why are we encoding ä when it is a combining diacritic but not ẽ? The fact that combining c-cedilla is a precedent doesn't make it any saner. For example a-breve is a separate letter in Romanian, and it is as well in the Atlasul linguistic romîn serie noua, just like the a-umlaut is in German dialectology. But in other dialectology works such as Atlas linguistique de la France, a-breve is just a breve a, not a different letter. You'll then end up with combining a-breve as a single diacritic for Romanian but combining a and combining breve for other languages. Yet the non-diacritic forms, i.e. regular letters would be represented by the same character sequence in Romanian or other languages, NFC ă or NFD ă. The same could be said for a-umlaut/diaeresis depending on how people are using it. Denis Moyogo Jacquerye
<<attachment: combining_a-breve-1-ALRsn.jpg>>
<<attachment: combining_a-breve-2-ALRsn.jpg>>
