On Mon, 2007-02-05 at 11:52 -0500, Tom Lane wrote:
> "Simon Riggs" <simon@2ndquadrant.com> writes:
> > Sounds like a good time to suggest making these values configurable,
> > within certain reasonable bounds to avoid bad behaviour.
>
> Actually, given what we've just learned --- namely that choosing these
> values at random is a bad idea --- I'd want to see a whole lot of
> positive evidence before adding such a configuration knob.
Sure. My understanding of the process we'd like to follow on this sort
of thing is:
1. make proposal, test for unseen negative effects or basic rejections
2. code performance prototype
3. assemble performance evidence
4. debate utility
5. complete coding
6. further review
Step 3 is always there for performance work, so even if you don't
mention it, I'll assume everybody wants to see that as soon as possible
before we progress.
-- Simon Riggs EnterpriseDB http://www.enterprisedb.com