FreeTDS Development Group

[Steve Langasek <vorlon AT netexpress.net>]

On Wed, Oct 10, 2001 at 11:39:20AM -0700, Curt Hagenlocher wrote:
> > Well, I maintain that UCS2 is broken, even if it's also used in Java.

> You'd have to define "broken" then.  An earlier spec of Unicode
> defined a 16-bit address space, and this is what both NT and
> Java have implemented.  At the time, it was the only spec that
> existed.

It is a widely accepted, and heavily marketed, belief that one of the
primary advantages of the Unicode standard is the unification of all
characters used in computing in a single, easy-to-implement character
set.  If you accept this, then any Unicode implementation built on UCS-2
is broken because it fails to meet the principal design criterion:
universality.  A 16-bit code space is not large enough for all human
languages for which textual electronic representation is desired.

You are correct that there were versions of the Unicode standard that
declared a 16-bit space for characters.  These standards were buggy, and
there are now versions of the Unicode standard which eliminate these
particular bugs.  Because after all, they were bugs:  if the standard
wasn't broken, would they have seen the need to fix it?

> This sounds like a gratuitous slag at Microsoft and, to a lesser
> extent, Sun.  Mind you, there are plenty of legitimate reasons
> to criticize these corporations, but their early and consistent
> support for a unifying standard that has since changed shouldn't
> be one of them.

Kudos to Microsoft for their early adoption of Unicode.  Slightly fewer
kudos to Sun, for I don't think they were uninfluenced by the fact that
using UCS-2 would make it a no-brainer to implement on Windows.  But
although they don't carry the blame for the brokenness of the standard
they implemented, this is not a reason for FreeTDS to expose such
brokenness to client applications.  FreeTDS is being written today, not
five years ago when UCS-2 was being put into NT4.  A Unix widechar
implementation done today should be using UCS-4.  Much better still is
the multibyte standard, UTF-8, which almost everyone in the Unix world
uses as the Unicode encoding of choice.

Steve Langasek
postmodern programmer

Attachment: pgpgBHJm1ynYX.pgp
Description: PGP signature