Name: bb33257 Date: 12/02/97


Currently, rules are built by adding to a list of default rules.
Commonly, additional characters for different locales are
inserted at 4 locations in the default rules:
after all the non-ignorables (NIs)
after all the primary ignorables (PIs)
after all the secondary ignorables (SIs)
after all the tertiary ignorables (TIs)

Unfortunately, you have to use specific characters to get this
effect currently. Thus changes in the default rules can ripple
through all the locales--creating potential for error.

To improve the robustness, add the additional syntax to the rules:

#N -- equivalant to &Z where Z is the last NI
#P -- equivalent to &\u2212 where \u2212 is the last PI
#S -- equivalent to &\u0020 where \u0020 is the last SI
#T -- equivalent to &\u009f where \u009f is the last TI

The implementation cost is trivial, and the '#' character is
among those reserved for these purposes.
======================================================================
###@###.### 11/2/04 18:15 GMT