On Thu, 23 Aug 2012 23:30:49 +0100 Michael Everson <[email protected]> wrote:
> On 23 Aug 2012, at 22:48, Jameson Quinn wrote: > > • We will have guessed wrong on the metadata, and the > > decision is made that a retroactive fix is the best solution, given > > that it (say) impacts only documents which mix Mongolian with Mayan > > numerals. > I don't know what you think a "retroactive fix" is but it is unlikely > that any "informative" as opposed to "normative" fix would be > possible. The immutable properties are: (i) name (ii) aliases (more can be added, but they cannot be removed) (iii) decomposition mapping (iv) canonical combining class (v) case folding (vi) case pairing (i.e. pairs and non-pairs of assigend characters remain such). Only (i) and (iv) seem plausible causes of problems. A canonical combining class of zero seems to be the preferred option for script encoding nowadays. Combining rules may have to be extended - that might be counted as an informative fix. > > • We will have guessed wrong on the metadata, but anyway > > need to encode multiple versions of the number glyphs, so having > > used up 20 codepoints for modern users of the Mayan numerals is no > > big loss. > We are not supposed to encode duplicate characters. We do where the properties necessitate, e.g. U+0241 LATIN CAPITAL LETTER GLOTTAL STOP and U+0294 LATIN LETTER GLOTTAL STOP, or the NEW TAI LUE and TAI THAM scripts. We also have the principal of the separation of scripts. Richard.
