Document PUA code points used by old decoders for old CJK(V?) encodings, as normalizing PUA artifacts is as important as Unicode normalization. Most of the problem should arise in Chinese and Han Nom characters, as there are more characters to screw up.
What if we write something that automatically generates bad jokes by substituting random words in a Wikipedia article for some boring:INPUT → funny:OUTPUT analogies? Word vectors can do that pretty well.
Infobox gene
There are some good stuff I can backport to here from the zh adaptions. Some messy "refactor" diffs coming up someday.
insert breaks to loops
replace ad-hoc string ops with not-very-ad-hoc ones (aliases, etc.)
probably go for a string table like CS1 is doing?
indents. let's face it we don't care about dirty diffs if it's fixed once and for all.
some styles like actual table literals.
and yeah we don't need to write chrTextTable out like that.
do early returns. why nest it when you can jump out of the wrong ones
Special:Search/"bundled in GNU coreutils" -POSIX: is part of the [[X/Open]] Portability Guide since issue 2 of 1987. It was inherited into the first version of POSIX and the [[Single Unix Specification]].<ref>{{man|cu|df|SUS}}</ref> It first appeared in <ref>{{man|1|df|FreeBSD}}</ref>
