Dave Cramer <pg(at)fastcrypt(dot)com> writes:
> Before we cut an 8.1 stable release, I'd like to revisit the use of
> unknown for strings at a minimum.
Isn't it *way* too late to be revisiting such questions for 8.1?
I'd be fairly worried about breaking cases that work conveniently
today. It sounds like something to experiment with during a devel
cycle, not to cram in at the last moment with no field testing.
regards, tom lane