"Simon Riggs" <simon@2ndquadrant.com> writes:
> That's true, but you don't know which one is the snapshot timestamp.
You're assuming there is such a thing as "the" unique active snapshot,
an assumption I find highly dubious. In any case, the reasons for
wanting to know which transactions are old have to do with the behavior
of VACUUM, and that only pays attention to the age of the whole
transaction not any individual snapshots. So I still don't see the
point of cluttering pg_stat_activity with yet more columns.
regards, tom lane
