Am Mittwoch, den 02.05.2012, 13:46 -0700 schrieb Doug Ewell: > Werner LEMBERG <wl at gnu dot org> wrote: > > >> So if two glyphs have enough "visual character" to be used in one > >> document to express two different meanings, then they should be > >> encoded as different characters? > > > > Yes, more or less. > > I don't think that's what was originally said. A quotation-mark > character that can be used as either an opening or closing mark, but > that doesn't kern correctly in some fonts when used as a closing mark, > does not seem to justify disunification.
Definitely not. Some kerning issues would just have been solved in passing had U+0022 and U+0027 logically been "disunified" and not only U +2018 and U+201C got their counterparts but also U+201A and U+201E. > > However, quotation characters need language > > tagging or something like that; you certainly don't want to have the > > situation to ask whether ' is the Byzantine opening quote, or ' the > > Martian alternate closing quote, or ' the you-name-it. It's a > > delicate issue. 'open' and 'close' have been "tagged to" U+0027 ('), encoding the "tagged versions" as U+2018 and U+2019; it only seems reasonable to do nothing less to U+002C (,) (and the same goes for adding the 'double' "tag"). Had they been restricted to use Courier, would they have said: "Hey, look, U+0027, U+2018 and U+2019 are the same. Let's unify them as, say, U+0027."? This has actually happened -- just in an inverted way: Had they been restricted to a font like TeXGyrePagella, everyone would have said that 'left opening high 6' (now encoded at U+2018) is not the same as 'right closing high 6', because the difference, which is the same as that between U+2018 and U+2019, would have been visible. But it wasn't visible. But such a potentially invisible difference was no reason not to "tag" 'minus' and 'en dash' to U+002D and encode the "tagged" versions as U +2212 and U+2013. > You don't want to turn this into a "German language" issue and have the > solution not work for quotational material in other languages that use > the same written conventions. Not at all. The issue of a missed "tagging" or disunification, or a wrong(?) unification, is *usage-specific* (and about logic, reason and consistency), *not language-specific*, though usage depends on language. Michael