On Thu, May 28, 2026 at 11:23 PM Jonathan S. Katz <jkatz@postgresql.org> wrote:
> Attached is the draft of the PostgreSQL 19 Beta 1 release announcement.
> Please review for accuracy (e.g. are features described correctly? Did I
> miss a revert), recommendations on additions or glaring omissions,
> structure, etc.
I asked Claude to analyze this for things that were missing. It
mentioned these as things that should possibly be mentioned: INSERT ON
CONFLICT SELECT, eager aggregation, IGNORE NULLS for window functions,
pg_dumpall for non-text formats. Of those, I would personally judge
eager aggregation to be the most significant.
It also pointed out that the bit about GRANTED BY is misleading,
because the syntax already existed, and already worked for roles.
What's different is that for non-roles, it now can be used with values
other than CURRENT_USER/CURRENT_ROLE and actually do something. It
suggested that this was actually too minor to be included in a release
note draft. I agree -- it's a small patch to fill a gap that few
people will ever really be troubled by. Of course it will be useful to
those who do run into it, but it's not really a headline feature IMHO.
It had various other complaints and suggestions too, of course, but
I'm not repeating them here because I didn't find them especially
interesting or entirely valid.
I do feel like this release announcement (like others you've written)
is extremely dense, and if I were writing it myself I would probably
mention somewhat fewer things so that the things I did mention hit
harder, but of course I'm not: you are. I also understand that gets
into tough arguments about what's actually worth mentioning, which are
inevitably subjective. Still, I'd suggest as a zero-order guideline
that including things that are <~ 500 LOC is more likely to be clutter
than a true headline feature.
--
Robert Haas
EDB: http://www.enterprisedb.com
