Philippe Verdy wrote: > It will be however simpler than sorting Thai with the > logical (phonetic) order, which requires a preprocessing > to find grapheme clusters and syllables with a dictionnary,
I've been told that that is not desirable, even though older dictionaries may have used such an ordering (maintained "by hand"). One reason it is not desirable may be that there are ambiguities. (Apart from ambiguities, this can be done without any prehandling, just inside the collation tailoring.) > unless you prefer to sort simply on the visual order I've been told that this is the desirable way to handle Thai ordering, except (of course) that "prevowels" are considered as being after the following consonant. (Logically/phonetically, it may belong later on, after a double, or rather "pair", consonant for instance.) UCA uses a prehandling to "swap" the prevowels and consonant, a 14651 tailoring for Thai and Lao can do the swapping in the collation tailoring itself (as is done in Linux locales for Thai and Lao already), to avoid the prehandling. See http://std.dkuug.dk/jtc1/sc22/wg20/docs/n1077-ThaiLao5.pdf. (I'm working on an update [shorter(!), using ".."-notation] to that paper as well. The preferred Lao ordering may be somewhat different, but I need confirmation(s!).) /kent k
