L2/06-067



Proposal to move some casing data from the UCD to the CLDR
Eric Muller
Adobe Systems Inc
February 8, 2006




Looking at the case transformation data expressed in the UCD, we find:

- for the vast majority of the cased characters, the mappings to their
  uppercase, lowercase, titlecase, and case-folded forms are to a
  single character and are unconditional

- a small number of cased characters have one or more of the mappings
  to sequences of characters rather than to a single character; those
  mappings are still unconditional

- U+03A3 GREEK CAPITAL LETTER SIGMA has a conditional treatment, recorded
  as a single entry

    03A3; 03C2; 03A3; 03A3; Final_Sigma; # GREEK CAPITAL LETTER SIGMA

- some mappings for orthographies which distinguish the dotted and
  dotless i and j; those conditional mappings involve at least a locale


The conditional mappings, which concern a very small number of
situations, make the description of the casing properties rather
complicated. If one views the lower case mapping as a code point (or
scalar value, or abstract character, or coded character) property,
then the value of that property is a mapping from a condition to a
sequence of code points. This is in sharp contrast with all the other
(non-Unihan) properties, which have simple values.

In fact, this complexity is not properly accounted for in the
standard. The Simple_Lowercase_Mapping property and its companions are
informally listed as "String properties" in PropertyAliases.txt. There
is also documentation of a Special_Case_Condition property as a
"String property", which does not make sense.

Furthermore, the tendancy in the Unicode Standard is to describe only
locale-independant data and processes (may be allowing for tailoring),
and to leave the handling of locale specific situations to other
mechanisms, such as the CLDR. Furthermore, it is arguable that the
final sigma behavior is actually conditional to locales based on the
Greek language, rather than inherent to the Greek script.


Given all this, I propose that:

- the conditional case mappings be treated as "tailorings" of the case
conversion operations, and that those tailorings be recorded in the
CLDR rather than the UCD;

- the property Special_Case_Condition be deprecated


---