Name: bb33257 Date: 09/24/98


RuleBasedCollator fails to build the Japanese collation sequence
using Unicode 2.1.2 database. The problem is that the decomposition
table (Normalizer) is decomposing math symbols in the Japanese
sequence into special symbols. These special symbols need to be
quoted to be understood by the parser. For example, \u2260 (Not
Equal To) is decomposed to \u003D (=) and \u0338 (overlay combining
mark).
The fix is to decompose the characters when the pattern entries
are built. This would require each pattern entry to call Normalizer.
decompose(). There may be performance issues involved.
======================================================================