L2/06-135 Source: Markus Scherer Date: 2006-04-17 16:26:36 -0700 Subject: Discourage adding new precomposed characters to CompositionExclusions Please add as a document and agenda item. (Bcc'ed to l2doc.) I propose that for character additions the UTC - explicitly consider decomposition mappings and composition exclusions - discourage adding new precomposed characters to CompositionExclusions, unless required by the stability policies. If a new character X is to be added to Unicode, and it is abstractly decomposable into characters Y+Z, where Y or Z (or both) is also a new Unicode character, then there are the following options that should be considered explicitly. The second part of my proposal means that option b) be discouraged. a) No decomposition mapping X is not canonically equivalent to Y+Z according to Unicode Normalization. b) X => Y+Z X decomposes to Y+Z under all normalization forms, that is, X always "disappears" under Unicode Normalization (X is added to CompositionExclusions.txt) c) X <=> Y+Z X decomposes to Y+Z under NFD and NFKD, and Y+Z compose to X under NFC and NFKC d) Unicode does not add X Y+Z is the only supported encoding e) Unicode adds X and Y but not Z, or adds X and Z but not Y No composition/decomposition possible. Note: This would require a change in the text in UAX #15, Section 6 Composition Exclusion Table, point 2. Post Composition Version. Sincerely, markus