Message-ID: <4532D17B.20106@modgraph-usa.com>
Date: Sun, 15 Oct 2006 17:25:31 -0700
From: "Craig A. James" <cjames@modgraph-usa.com>
User-Agent: Thunderbird 1.5.0.7 (X11/20060909)
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: Hints proposal
References: <78ED28FACE63744386D68D8A9D1CF5D4209CBF@MAIL.corp.lumeta.com>
In-Reply-To: <78ED28FACE63744386D68D8A9D1CF5D4209CBF@MAIL.corp.lumeta.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

So let's cut to the bone: If someone thinks a proposal is a bad idea, and they're volunteering their time on an open-source project, why would they implement the proposal?

In all the heat and smoke, I believe there are two basic conclusions we all agree on.

1. Optimizer:
   a) A perfect optimizer would be a wonderful thing
   b) Optimization is a hard problem
   c) Any problem that can be solve by improving the optimizer *should*
      be solved by improving the optimizer.

2. Hints
   a) On a aesthetic/theoretical level, hints suck.  They're ugly and rude
   b) On a practical level, introducing hints will cause short- and long-term problems
   c) Hints would help DBAs solve urgent problems for which there is no other solution

The disagreements revolve around the degree to which 1 conflicts with 2.

1. Developers feel very strongly about 2(a) and 2(b).
2. DBAs "in the trenches" feel very strongly about 2(c).

So my question is: Is there any argument that can be made to persuade those of you who are volunteering your time on the optimizer to even consider a HINTS proposal?  Has all this discussion changed your perspective on 2(c), and why it really matters to some of us?  Are we just wasting our time, or is this a fruitful discussion?

Thanks,
Craig