Stephan Szabo <sszabo@megazone.bigpanda.com> writes:
> Potentially on some systems it'd be possible to
> generate a case insensitive collation as part of a locale
> and then use such for LC_COLLATE on initdb which would make all
> comparisons of text fields case insensitive.
Another possibility is to create a 'case insensitive text' datatype
that has its own case-insensitive comparison operators and associated
btree opclass. You would want to create an implicit cast to text and
an automatic cast from text so as to take advantage of the existing
text-type operations. I'm not sure whether this would work cleanly
or would result in can't-choose-the-operator-to-use errors, but it'd
be worth trying.
regards, tom lane