Subject: Re: [HACKERS] Slow count(*) again...
From: Jeff Davis <pgsql@j-davis.com>
To: Mladen Gogala <mladen.gogala@vmsinfo.com>
Cc: Chris Browne <cbbrowne@acm.org>,
	"pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
In-Reply-To: <4D4B231C.3030802@vmsinfo.com>
References: <201102012247.p11Ml6u02682@momjian.us>
	<4D48914B.4050403@dunslane.net> <4D489AD1.1010105@vmsinfo.com>
	<AANLkTiknZi97dDwm7sHqy0EL7tpNEP9sY5uPV6AN+q=u@mail.gmail.com>
	<4D499E55.4020107@vmsinfo.com> <4D49A6B9.4060608@2ndquadrant.com>
	<4D49C482.4010707@vmsinfo.com> <4D49F0BA.9000901@2ndquadrant.com>
	<4D4A16BF.1020304@vmsinfo.com> <4D4A4844.3020409@2ndquadrant.com>
	<4D4AD9F6.6040903@vmsinfo.com> <4D4AE16E.9040807@peak6.com>
	<4D4B17B4.9030406@vmsinfo.com>
	<871v3o3jgu.fsf@cbbrowne.afilias-int.info>
	<4D4B231C.3030802@vmsinfo.com>
Content-Type: text/plain; charset="ISO-8859-1"
Date: Thu, 03 Feb 2011 19:01:10 -0800
Message-ID: <1296788470.16666.119.camel@jdavis-ux.asterdata.local>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit

On Thu, 2011-02-03 at 16:50 -0500, Mladen Gogala wrote:
> Chris Browne wrote:
> > Well, the community declines to add hints until there is actual
> > consensus on a good way to add hints.
> >   
> OK. That's another matter entirely.   Who should make that decision? Is 
> there a committee or a person who would be capable of making that decision?

Admittedly I haven't read this whole discussion, but it seems like
"hints" might be too poorly defined right now.

If by "hints" we mean some mechanism to influence the planner in a more
fine-grained way, I could imagine that some proposal along those lines
might gain significant support.

But, as always, it depends on the content and quality of the proposal
more than the title. If someone has thoughtful proposal that tries to
balance things like:
* DBA control versus query changes/comments
* compatibility across versions versus finer plan control
* allowing the existing optimizer to optimize portions of the
  query while controlling other portions
* indicating costs and cardinalities versus plans directly

I am confident that such a proposal will gain traction among the
community as a whole.

However, a series proposals for individual hacks for specific purposes
will probably be rejected. I am in no way implying that you are
approaching it this way -- I am just trying to characterize an approach
that won't make progress.

Regards,
	Jeff Davis