2013/2/3 Costello, Roger L. <[email protected]>: > - It is easier to use a few keystrokes for combining accents than to set up > compose key sequences for all the possible composed characters.
But MOST texts using combining diacritics are written in languages for which there already exists standard keyboards featuring combine keys or dead keys. These standard keyboards are handy and do not require lot of compose key sequences, and they are avialable by default "out of the box" in almost all OSes. All these keyboards will compose letters in precombined forms, which are already NFC for most Latin/Cyrillic/Greek. The exception is Vietnamese for which there are several competing keyboards allowing inputs in various forms (not always normalized, but most often partly precombined at least by pairs). Hebrew is most often typed and encoded in precombined form (for the most frequent diacritics, notably SIN and SHIN dots). Sometimes anyway you'll find characters that are not in any standard normalization form (neither NFC nor NFD). So you cannot recommand any form. But if the W3C needs to update something, it's to say that ALL forms that are canonically equivalent should be treated equally. This means that it is to the recipient of encoded documents to perform their own normalization. But there's no recommandation about which normalization will be used by the recipient : for rendering NFC is generally easier, for searching NFD may be better but the more effective way for searching is to use collation, which already includes a NFD step in the standardized algotithm, but collation can also be written which does not require a prior normalization when it is not needed). Unicode also decribes some properties that allows processing with "fast normalisation checks", this can be used for creating conforming collators.
