Hi,
On 2018-08-25 14:47:20 -0400, Bruce Momjian wrote:
> Well, our normal logical is whether the average user would adjust their
> behavior based on this change, or whether it is user visible. I thought
> it was a contrived-enough query that this was not the case, but I would
> be interested to hear what others think.
I think that's less "our" logic and more yours, that has become
established because you've done most of the major release notes for a
long time. I'm not trying to say that that's wrong or anything, just
that a lot of those principles grew organically without explicit
decisions.
I also think that the need of userbase have shifted - it used to be that
a lot more people couldn't migrate to PG for lack of features, but these
days we have a much larger userbase that has to migrate from version to
version on a regular basis.
Planner changes, in my experience, are *the* major hurdle in doing so
(leaving aside things like the 8.3 casting changes), because they have
the potential to drastically change performance. In a largely
unpredictable way. So one of the first things you do after encountering
an issue while testing an issue during a migration, is to look through
the release notes whether that query could be affected by a documented
change.
Greetings,
Andres Freund