Hiroshi Inoue writes:
> For example, there could be case-insensitive collations.
You could do that, but normally you would take account the case
differences in the third pass.
Another interesting problem is how you would expect to handle
multicharacter collating elements that cross wildcard boundaries (like
'oc%' when 'ch' is a collating element). That would give weird results.
--
Peter Eisentraut peter_e(at)gmx(dot)net