Re: [pgsql-advocacy] Assembling "top features" list for betaannouncement - Mailing list pgsql-advocacy

JD, Robert,

* Joshua D. Drake (jd@commandprompt.com) wrote:
> On 04/09/2017 02:56 PM, Robert Haas wrote:
> >>3-5 is the sweet spot for a PR. I think SCRAM being listed is a very good
> >>thing.
> >
> >Well, the problem is that everybody's going to have their own ideas on
> >what is most important.
>
> That is definitely true but I think that is where leadership comes
> in. What do our primary users think is important?

I don't actually know that it's our existing users that we want to be
targeting with this.  Further, I'm not sure I follow what you mean by
"primary users."  Did you intend "majority of?"

> Better Partitioning and Logical replication are a given. Parallel
> query (AWESOME!) and SCRAM are probably at a head based on the
> audience.

I prefer the term used elsewhere that I saw- "Native" partitioning,
instead of "Better."  In line with the idea that we really want to
target people who don't currently use PG to reconsider why they don't
and if they should, things like "Native partitioning" and "logical
replication" are big, new, features which may make them consider PG more
seriously than they did before.

> It is true that some will care about better query planning with RLS
> but it is a *very* small subset of our user base.

I agree that a minority of existing users are interested in RLS.  In
addition- we already have it and made a splash with it being released as
part of 9.5.  Improvements to it are certainly note-worthy, but talk of
improved planning with RLS strikes me as less likely to make non-PG
users consider switching to PG.  To some extent, the same goes for
improved parallel query, though the situation is a bit different there
since it wasn't enabled initially either.

> >- we can't list everything - but we're blessed (or cursed) with
> >needing to promote a release that really has a ton of great stuff in
> >it, and we don't want people to tune out because the stuff they care
> >about isn't listed.
>
> True but consider this: The people who care about better query
> planning with RLS are already watching the lists for better query
> planning with RLS.

I'm not sure that's really accurate.  I'm sure some are, but not all of
them.  That said, I agree that it's still a minority of our overall
userbase.

> From an Adoption and "What is going to hit the presses" perspective,
> things like Replication and easy(er) Partitioning are going to hit
> more hearts and minds.
>
> From an adoption perspective the question we should ask is: "What
> are the people running RDS, Compose Postgres and Heroku Postgres
> going to care about?. The top three there are, "Better Partitioning,
> Logical Replication (Logical replication allows you to replicate OUT
> of RDS) and Parallel query with a very close combat of SCRAM."

I don't think we should focus only on what users of hosted platforms are
interested in.  For one thing, those users are already using PG and will
be upgrading based on what the hosted platform provides, or forces on
them.

> Parallel Query, Scram and better query planning with RLS are very
> much enterprise features (with Parallel query being the most cross
> culture).

I agree, and we should certainly be considering those users.  Put
another way, I'm worried that the focus here is on the number of PG
installs (hosted platforms have tons) without consideration for the
other end of the spectrum (large, locally hosted, systems).  Both are
important to us.  I don't have a great answer to which is *more* so (one
could start to consider "brand awareness" vs. "contribution to the
community" and other such metrics, but we'd then have to determine what
our goals for this are, exactly, and this thread doesn't strike me as
the right place to try and hash *that* out..).

Thanks!

Stephen