Martijn van Oosterhout <kleptog(at)svana(dot)org> writes:
> The issue is whether anything you want to ORDER BY needs to be
> described by an B-tree operator class, and hence have a real sort
> function.
> 1. Do people have any problems with this?
> 2. Would a patch for this be accepted seperate from the whole collation
> stuff?
I think it's reasonable to remove that feature, *after* we provide
a workable substitute. So, "no" to both questions ...
regards, tom lane