Re: Rework the way multixact truncations work

On Thu, Dec 10, 2015 at 08:55:54AM -0500, Robert Haas wrote:
> I don't know have anything to add to what others have said in response
> to this point, except this: the whole point of using a source code
> management system is to tell you what changed and when.  What you are
> proposing to do makes it unusable for that purpose.

Based on your comments, I'm calling the patch series returned with feedback.
I built the series around the goal of making history maximally reviewable for
persons not insiders to commit 4f627f8.  Having spent 90% of my 2015
PostgreSQL contribution time finding or fixing committed defects, my judgment
of how best to achieve that is no shout from the peanut gallery.  (Neither is
your judgment.)  In particular, I had in view two works, RLS and pg_audit,
that used the post-commit repair strategy you've advocated.  But you gave me a
fair chance to make the case, and you stayed convinced that my repairs oppose
my goal.  I can now follow your development of that belief, which is enough.