Name: mf23781 Date: 01/16/98


Actually raised against JDK 1.1.6

JIS, CJK collators could be built algorithmically

Currently, the collation rule tables for Japanese and Chinese
are huge, because they contain entries for lots of ideographic
characters. Since these tables are part of the locale data,
it takes a very long time to load locales that uses these
languages.

In most collators, the ideographs are sorted in the same
order as they appear in the standard national character set,
and the JDK already contains character converters that know
all about the ordering of the national character sets. It
would be a huge performance win if we were to build the
collators from the character converters' data rather than
maintaining a separate table.

Again this is performance work Taligent plan to do eventually.
======================================================================