L2/04-167

Date/Time:    Mon Apr 12 19:26:21 EDT 2004
Contact:      Asmus Freytag
Report Type:  Error Report
Subject: Inconsistency between data files in 4.0.1

In 4.0.1 an inconsistency has been introduced in the tags for certain
properties:

1) compatibility decomposition tags used to be all lower case.
From 4.0.1 these are Titlecased in PropertyValueAliases.txt and
DerivedCompatibilityTags.txt while still lowercase in
UnicodeDatat.txt and NamesList.txt

2) the joining group tags used to be all upper case
From 4.0.1 these are Titlecased in PropertyValueAliases.txt and
DerivedJoiningGroup.txt while still uppercase in
ArabicShaping.txt

While the casing differences do not affect the identity of these
tags, the change does affect software.

We allow the use of property tags in *implementations* in all casing
variations and with spaces, hyphens or underscores inserted,
in order to make it possible to create identifiers from them,
even in contexts where specific rules or conventions
proscribe or prohibit certain styles.

However, that's not the same as changing the casing etc. of these
in the database files. Doing that requires every piece of sotware
that tries to correlate the various files to implement this form
of ad-hoc loose matching.

Having an inconsistent spelling between the datafiles and the
property/value aliases makes it unnecessarily difficult to
use simple tools like grep to look up all occurrences of a
given tag across the UCD. It also breaks some other software.

I noticed this when my character
browser, http://www.unicode.org/unibook
was no longer able to read Arabic Shaping data from the datafile.
There may be other software out there, whether publicly
distributed as my
viewer, or for internal use that runs into the same issue.

This change makes creating diffs of the derived files useless
unless you happen to use a case-insensitive tool.

This particular casing should be considered an erratum and the casing
restored for 4.1 to its original values.