>I think the citext extension[1] already made much of this straightforward? sort of. It was doing case-folding based on LC_CTYPE set at initdb time, so if you

>I think the citext extension[1] already made much of this straightforward? sort of. It was doing case-folding based on LC_CTYPE set at initdb time, so if you have data of different locales to deal with, it would only work for one locale and be (subtly) broken for others (Turkish comes to mind) It's also based on glibc's interpretation of case-folding rather than what the Unicode consortium has agreed upon. Depending on locale and tolerance to weirdness that can be anywhere between a non-issue and a show-stopper. Case-folding is very tricky business.