Re: 9.0: Too many features. Help us choose!

On Tue, Jun 22, 2010 at 19:16, Josh Berkus <josh(at)agliodbs(dot)com> wrote:
>
>> I don't remember what I put in for that, but here's how I thought on a
>> number of cases. The LISTEN/NOTIFY improvements are important to
>> people who have been using postgresql for a long time, and use it in a
>> way that's not all that common these days (look, ma, no ORM!). For an
>> *outsider*, it's completely irrelevant - they didn't know there was a
>> problem before (unlike vacuum which people have heard of forever,
>> nobody has heard of issues with listen/notify), so it looks more like
>> trying to push something because we didn't have enough relevant.
>
> Yeah, I guess my perspective is different.  From where I sit, LISTEN/NOTIFY
> was a useless feature before: it didn't carry messages, it had severe
> performance limitations.  Suddenly, with the overhaul
> , PostgreSQL has built-in transactional message queueing.

Given the large number of installations I've come across that have
been very happy with the previous LISTEN/NOTIFY, it's *far* from being
useless in the versions we have out there today. There's certainly
been room for improvement - which we now have - but calling it useless
just tells me you're far disconnected from reality :P

> In other words, previously most people were unaware that LISTEN/NOTIFY
> existed and I wouldn't have recommended it to them.  Now it's a useful tool
> which users can use to build new kinds of applications.  HStore is the same
> ... we had it in 8.4, but it wasn't useful.  Now you can build an
> application around it, as long as you don't use "=>".
>
> Contrast this with the additional windowing functions, which received a
> about an equal number of votes.  In my experience, most of the public
> doesn't even know what windowing functions are, and could care less that we
> implemented 5 more of them.  As far as the casual user is concerned, we
> implemented windowing functions in 8.4 and we're done now.

If they come from MySQL, they won't.

If they come from one of the big databases, they will.

The simple ability to do a moving average has a lot more users amongst
my customers than the changes in LISTEN/NOTIFY, really.

> From my perspective, the press release should focus on features which answer
> the question "Why would I use PostgreSQL instead of another databse?".  I
> think that people here on the list agree in principle but nevertheless tend
> to focus on features which support incremental improvements of existing
> functionality over features which support entirely new applications, if
> about half the votes are anything to go by.

Uh, the LISTEN/NOTIFY change is as much an incremental improvement as
the windowing functions, to keep up with your examples. You could do
them both previously, but neither solution really scaled.

> Anyway, the voting did let me get a list of 5 "don't bother" features and 5
> "must have" features, which then means that the rest can be based on PR
> discussion, and is what I expected.  And let me settle the release notes,
> where our space constraints are less.
>
> The only question is ... should I broadcast this survey on -general to try
> to get the perspective of more casual PostgreSQL users?  Over half of the
> current survey respondants called themselves "Experienced PostgresQL Users"
> which is, I think, where part of the voting skew comes from.

Probably not a bad idea.

--
Magnus Hagander
Me: http://www.hagander.net/
Work: http://www.redpill-linpro.com/