Peter Geoghegan <peter.geoghegan86@gmail.com> writes:
> I think we should bite the bullet and adopt ICU,
I see absolutely nothing to recommend that course of action. Reasons not
to:
1. Being compatible with the operating system's collation behavior is a
feature, not a bug. If nothing else, it allows us to tell people that
if we sort data the same way that sort(1) does, then it's not a bug that
we're not sorting the way they think we should. But quite aside from
that, there are practical uses to being compatible with other tools.
2. Last I checked, ICU *only* supports Unicode, and not only that, but
only UTF16. This is a non-starter; not only for our Far Eastern users,
but also those who find various LatinX encodings sufficient. ICU would be
a functional fail for the former and a performance fail for the latter.
3. As Thomas Munro already noted, whatcha gonna do when ICU changes their
collations? Or are their collations graven on stone tablets, unlike
anyone else's?
We certainly could stand to put some work into the problem of coping with
collation changes. But claiming that ICU is the solution, or even a
solution, seems obviously wrong.
regards, tom lane