Re: Unicode FFFF Special Codepoint should always collate high.

On Thu, Jun 24, 2021 at 10:29 AM Thomas Munro <thomas.munro@gmail.com> wrote:
> On Wed, Jun 23, 2021 at 9:57 PM Telford Tendys <psql@lnx-bsp.net> wrote:
> > I trust those guys, they will figure it out. I strongly predict that
> > they will keep the behaviour consistent with RHEL 7.
>
> I'd doubt that.  It's well known that glibc 2.28 (what RHEL8 upgraded
> to) included changes that affected everybody by changing the sort
> order of common symbols like '-' (though every upgrade potentially
> contains subtle changes affecting just a few specific languages), but
> I consider the recent big change an improvement because it now agrees
> more often with other operating systems and libraries that use CLDR.
> Even if you are right that FFFF's sort-high rule should be exposed to
> users (need references), RHEL7 was also wrong in that case.

Oh (following along with your bug report)... so glibc 2.28+ is based
on ISO 14651, not UCA/CLDR.  They are related but different[1] (I was
confused about that).  That rule is coming from a CLDR document.  So
one question is whether anything similar is in the ISO document[2].

/me wanders away wondering if an OS that *is* using CLDR is supposed
to be collating \uFFFF the way ICU does

[1] https://unicode.org/reports/tr10/tr10-34.html#Synch_ISO14651
[2] https://standards.iso.org/ittf/PubliclyAvailableStandards/index.html