On 2025-08-26 13:45, Paul Eggert wrote:
> So it looks like we should avoid <q>.

After reviewing the comments on this topic it seems that plain UTF-8 is
the way to go. I went through the HTML and public-facing files like
README (visible on GitHub), and used UTF-8 quotes and a few other
special characters to make the files easier to edit and/or read. The
complete set of non-ASCII characters that now occur directly in *.html,
*.txt, or other public-facing text files is now:

   §«°±»½¾×–‘’“”•→−≤★⯪

and this set is now documented and checked in Makefile as
UNUSUAL_OK_CHARSET. Although we can extend this character set as needed,
I thought it better to not allow arbitrary UTF-8 due to the usual
problems with normalization, confusables, etc.

I installed the attached set of proposed patches to implement this. The
first patch is mostly a mistake but that is rectified in the last patch.

I considered going further, by using curved quotes and other special
characters in commentary in the main (non public facing) data files,
e.g., to say “34° 54′ S” instead of the current “34° 54' S” so that
comments use U+2032 PRIME instead of the approximation U+0027 APOSTROPHE
to denote minutes. However, this would be more work and I was not sure
it would be worth the maintenance hassle long term so I left it alone
for now.

The only change these patches make to data (as opposed to commentary)
consists of changing Côte d'Ivoire to Côte d’Ivoire in iso3166.tab. The
patches also change Gur'yev to Gur’yev and Dumont d'Urville to Dumont
d’Urville in the comments column of zone1970.tab, and remove some
unnecessary "s in the comments column of zonenow.tab.