L2/11-003


Title: An Alias for Mark

Source: Ken Whistler

Date: January 11, 2011

Action: For review by UTC


Background

There is a small inconsistency between the list of combined
General Category aliases that we have formally defined
in PropertyValueAliases.txt and their actual usage in other
contexts.

The combined aliases do not represent primitive property values
for General Category, but instead certain useful unions of
those property values. Most are single-character aliases for
the major character categories that General Category subdivides
into further subtypes. The exact list of these currently defined
in PropertyValueAliases.txt is:

gc ; C         ; Other                    # Cc | Cf | Cn | Co | Cs
gc ; L         ; Letter                   # Ll | Lm | Lo | Lt | Lu
gc ; LC        ; Cased_Letter             # Ll | Lt | Lu
gc ; M         ; Mark                     # Mc | Me | Mn
gc ; N         ; Number                   # Nd | Nl | No
gc ; P         ; Punctuation  ; punct     # Pc | Pd | Pe | Pf | Pi | Po | Ps
gc ; S         ; Symbol                   # Sc | Sk | Sm | So
gc ; Z         ; Separator                # Zl | Zp | Zs

The problem arises from the fact that for gc=M, our usual practice
is to refer to "Combining Mark", rather than just "Mark". Ordinarily
this would be just a curiosity, but there is a formal issue related
to our statement of normative definitions in Chapter 3 of the standard.

Section 3.6 Combination, D52 defines "combining character" as follows:

  D52 Combining character: A character with the General Category of
      Combining Mark (M).
      
And D50 and D51 use similar wording.

The problem is that we don't actually define "Combining Mark" as a formal
value for General Category anywhere -- all we define is the alias "Mark".

Rather than touch the normative definitions in Chapter 3 to fix this,
I think the easiest and most obvious fix is to simply add "Combining Mark"
as another alias for gc=M in PropertyValueAliases.txt.

There is an additional reason to do this now, because the IDNA documentation
is somewhat sloppy about its use of terms "combining character" and
"combining mark", creating some confusion for IDNA implementers. The
fact that our own formal definition of "combining character" uses
"combining mark" as an undefined term hasn't been helping.

Proposal

Add "Combining Mark" as a second alias for gc=M in PropertyValueAliases.txt
for Unicode 6.1.

Additional Issue for UAX #44

The set of combined values for General Category is not adequately documented
anywhere outside of PropertyValueAliases.txt itself. Nor is it clear to
everyone that their status is distinct from the basic General Category
values, since they represent convenient aggregations of unions of values,
rather than simple values on their own.

To address this I suggest that for Unicode 6.1 we add relevant documentation
to UAX #44, in Section 5.7.1 General Category Values, explaining the
distinction clearly, and then adding an explicit table of the combined
values that are actually defined in PropertyValueAliases.txt. That will
make it easier for implementers of regex engines and the like to understand
these values.

.