Bruce Momjian <pgman@candle.pha.pa.us> writes:
> VACUUM FREEZE. Interesting idea. Did we have that in 7.2? I never
> thought of using it. Good idea.
IIRC, it was new in 7.2 --- but pg_upgrade goes back further than that.
I am not sure if this idea just escaped us before, or if there's a hole
in it. We need to study it carefully.
> Why do we have to do WAL? Do we have WAL log id's in the tuple headers?
Not in tuple headers, but in page headers. The buffer manager will
choke if the LSN fields point past the end of WAL.
regards, tom lane