Re: Big disappointment with Postgres

From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Sun, 6 Feb 2011 19:28:09 +0000 (UTC)
Message-ID: <pan.2011.02.06.19.28.09_at_gmail.com>



On Sun, 06 Feb 2011 05:09:07 -0800, prunoki wrote:

> On Feb 6, 6:55 am, Noons <wizofo..._at_yahoo.com.au> wrote:

>> Mladen Gogala wrote,on my timestamp of 6/02/2011 9:13 AM:
>>
>> > On Sat, 05 Feb 2011 23:38:08 +1100, Noons wrote:
>>
>> >> Beg your pardon?  So, a hint that fixes a plan "introduces upgrade
>> >> and maintenance issues", but a patch to fix a specific problem with
>> >> the optimizer does not introduce such?
>> >> Hellooooooo?
>>
>> > I posted that here because Postgres has recently been advertised as
>> > an alternative to Oracle. It's nothing of the kind and the people
>> > ogling it should be aware of that. It was foolish of me to expect
>> > otherwise. It was an off topic post which otherwise doesn't belong on
>> > this group, but that was the reason for me to put it here.
>>
>> It's good to have this sort of thing exposed.  These morons rely too
>> much on their little forums where everyone jumps the same way when the
>> whistle sounds. They need to be shown clearly that they are a vast
>> minority, unlike what their cozy little corners lead them to believe.

>
> Hmmm, not giving me the tools to let me solve my own problems in the way
> I want seems like a bad idea. And yes, migration projects can be
> extremely tricky that way. There is time for evil and dirty to make it
> work and there is time to remove them later.

There are two excellent papers here:
http://www.singingsql.com/WhyGoodOptimizersDoNeedaHintLongerVersion.htm http://www.nocoug.org/download/2005-08/
WhenGoodOptimizersMakeBadChoices.doc

Both were written by Dan Tow, a long term DBA who has my utmost respect. The papers pre-date this debate for several years and present a very coherent argument in support of hints.
They do not cover the psychological side of things. Telling users that PostgreSQL creators consider them dumber than a computer program is simply not a good idea if PostgreSQL is supposed to be an alternative to Oracle. It isn't. I am glad that we've ironed that out.

-- 
http://mgogala.byethost5.com
Received on Sun Feb 06 2011 - 13:28:09 CST

Original text of this message