🚀 go-pugleaf

Submitted-by: gwyn@smoke.brl.mil (Doug Gwyn)

In article <16h5a2INNko4@ftp.UU.NET> peterw@spaten.sharebase.com (Peter Wisnovsky) writes:
>... my impression of what the potential problem is that if the standards
>orgs went ahead and used wchar_t as the vehicle for supporting Unicode,
>then existing conformant implementations that define wchar_t to be an
>8-bit or 32-bit value would be made non-conformant. NT is not
>non-conformant in itself...but the solution it proposes to the storage
>of Unicode data, that is, the fixing of the size of wchar_t to be
>16=bits wide, would not work with existing conformant systems.

This represents a misunderstanding of the role of standards and of wchar_t
in particular.  The C standard requires that a conforming implementation
define a wchar_t type for the internal representation of whatever
multibyte character sequences it chooses to support; however, it does not
mandate that any particular multibyte encoding be supported.  A vendor may
conform to the C standard while not supporting Unicode, for example.  It
is left as an implementation "marketing decision" just how extensive the
multibyte support will be.  Certainly, an implementation that chose to
ignore genuine extended character sets and define wchar_t as an 8-bit
type will not be able to at the same time support 16-bit Unicode.  If and
when such implementations are revised to properly support extended
character sets, they will have to make wchar_t at least 16 bits, as most
international implementations of Standard C already do.  Since wchar_t is
an internal data type it is not really relevant to issues of data
interchange among different systems; that does not lie within the scope
of the programming language standard.

>Two proposals were discussed at the Unicode conference wrt XPG. One
>would be to have the standards changed so that wchar_t would be
>defined to be 16-bits wide; the other would be to create a new
>datatype, `unichar'. The sentiment at the conference was to create a
>new datatype.

If that accurately reflects the discussion, then it merely serves to
confirm the widespread impression that the Unicode proponents don't
understand the existing standards nor the magnitude of the effects of
addition of new data types to existing languages.  It is already an
easy matter to, for purposes of conformance with other standards, add
non-conflicting requirements (such as at least 16 bits in wchar_t)
beyond base standards.  For example, POSIX.1 adds library requirements
beyond those specified in the C standard.  There is no need to "change"
the base standards in such cases.

Standard C's multibyte character support was designed in close
consultation with many individuals and organizations who had long been
involved in "internationalization" issues.  ITSCJ particularly comes
to mind, and they have continued to work on improvements within the C
multibyte character model.  Originally they too suggested a separate
data type for "long" character encodings, and I proposed an alternate
suggestion that also introduced a new data type (my type would have
been used for sub-character bytes, however, reserving "char" for the
sole character type, thus immensely simplifying programming).  After
considerable debate and several committee and working subgroup meetings,
consensus was reached on the multibyte external sequence/wchar_t
internal encoding approach.  It would behoove the Unicode proponents to
fully understand those deliberations and the resulting design before
they further bollix up the works.


Volume-Number: Volume 29, Number 7