On Wed, 18 Jan 2017 13:35:55 -0700 "Doug Ewell" <[email protected]> wrote:
> Richard Wordingham wrote: > > > I think it is not a 'typographical error' if it renders as it > > should! > > What if it renders correctly on some systems but not on others? > I do see your point, though. Writing systems that permit different > spellings of the same glyph (cluster), only one of which is 'correct' > even after normalization, can be tricky like this. I think this would > still be a matter of 'misspelling' rather than 'miscoding' because a > typist should not have to be concerned with character codes per se. As you've put it, it sounds like the way things were with a simple Thai typewriter. A vowel below, a vowel above and a tone mark could be typed in any order, as though they had three different non-zero combining classes. Thais were trained to type into computers by input routines only accepting the marks in the correct order - this was before the days of canonical combining classes. In the case of greatest concern to me, there can be two different orders, but only one is appropriate for a given word. In most cases, only one word of that appearance exists, and one can usually guess which one does exist. (That is why the system works despite the occasional ambiguity.) It's not unlike how Thai would work had phonetic order been successfully insisted upon, except that there is no evidence that sorting should be by appearance, whereas in Thai as it was encoded before Unicode (and is now, after normalisation), encoding and sorting are based purely on appearance. (Well, officially - in practice, Thais appear to sort by doing syllable-by-syllable comparisons.) In this case of concern, the range of renderings is occasionally different, which is another reason that two different encodings for the same appearance must be tolerated. Richard.
