The background document for PRI #308 (Property Change for NNBSP), http://www.unicode.org/review/pri308/pri308-background.html , says,
"The only other widely noted use for U+202F NNBSP is for representation of the thin non-breaking space (espace fine insécable) regularly seen next to certain punctuation marks in French style typography. However, the word segmentation change for U+202F should have no impact in that context, as ExtendNumLet is explicitly for preventing breaks between letters, but does not prevent the identification of word boundaries next to punctuation marks." Unfortunately, this isn't quite true. In the text fragment " dit<NNBSP>: ", there would be internal word-boundaries before 'd' and before and after ':', but the word isolated would be the four characters "dit<NNBSP>". One solution would be replace NNBSP by U+2009 THIN SPACE, for with untailored line-breaking there would be no line break between it and the 't' or colon, but there would be a word break between the 't' and the thin space. The problem is that characters with property ExtendNumLet can be the first or last character of a word as well as a character strictly within a word. In this respect, the property differs from characters with the property MidNumLet. The problem with using that property instead is that such characters, such as FULL STOP, may be flanked by letters or numbers within a word, but not both. The problem then arises with the Mongolian analogue of '4th' etc. - it is written digit, NNBSP, letters, and is a single word. Richard.
