On Thu, Feb 3, 2011 at 11:56 AM, Bruce Momjian <bruce@momjian.us> wrote:
> The settings are currently there to better model the real world
> (random_page_cost), or for testing (enable_seqscan). They are not there
> to force certain plans. They can be used for that, but that is not
> their purpose and they would not have been added if that was their
> purpose.
Sure. But Mladen's point is that this is rather narrow-minded. I
happen to agree. We are not building an ivory tower. We are building
a program that real people will use to solve real problems, and it is
not our job to artificially prevent them from achieving their
objectives so that we remain motivated to improve future versions of
the code.
I don't, however, agree with his contention that this is easy to
implement. It would be easy to implement something that sucked. It
would be hard to implement something that actually helped in the cases
where the existing settings aren't already sufficient.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
