L2/02-035

I propose that we make the following addition to the end of 3.6 Decomposition (revision) in Unicode 3.2.

Add the following text after D23:

D23a Canonical composite: a character which is not identical to its canonical decomposition.  

• For example: U+00E0 LATIN SMALL LETTER A WITH GRAVE is a canonical composite, since its canonical decomposition is to the two characters: a and grave. U+212A KELVIN SIGN is a canonical composite since its canonical decomposition is to a K.
• Using D21 and D23a, all characters can thus be broken down into the following four categories:
  
  

  Category	Examples
  non-composite	a, b, c, ...
  canonical composite	a-grave (à), kelvin sign (K), ...
  compatibility composite	micro-sign (µ), square mil (㏕), ...
  canonical & compatibility composite	greek upsilon with acute and hook symbol (ϓ), ...

Background

We define decomposable character in D18. We say that it is equivalent to one or more characters according to the decomposition mappings. We give composite and precomposed as aliases for this term. However, we do not, in that definition, distinguish between the kinds of decompositions that determine a decomposible/composite.

To resolve an ambiguity in our terminology, in 3.2 we are using the term compatibility composite in D21. That distinguishes a particular subset of composites, based on the kind of decomposition. The term canonical composite distinguishes a different subset of composites, based upon the other kind of decomposition. While it is in some sense is a natural fallout from the definitions of canonical and composite, it is important enough that we should have a formal definition.

People too easily misuse the term composite (or precomposed) to mean only canonical composite -- I've seen this on a number of occasions. A formal definition -- and inclusion of the table with examples -- will help to reduce that confusion.