Re: Big disappointment with Postgres
Date: Sat, 5 Feb 2011 14:21:40 -0000
"Thomas Kellerer" <OTPXDAJCSJVU_at_spammotel.com> wrote in message
> Tom Kyte strongly detests hints as well:
> Q: When should hints be used:
> A: Never. They are the path of last resort.
"Never" and "last resort" aren't mutually compatible. I think we can infer from this that "never" is an exaggeration for emphsis
> "But -- I also have a rule -- don't use HINTS."
But the next sentence lists some hints that Tom thinks are okay - so there's another citation that isn't the black and white decision that you imply..
> "If you find you are hinting every other query in your system --
> something is
> obviously wrong and we need to fix it. Abusing hints is not recommended,
> you are just
> building another RBO if you do that -- precluding the software from doing
> its job. Might
> as well not have an optimizer at that point."
"every other query", "abusing hints" - I think we get the idea from this that Tom's complain matches the fears of the Postgres developers, that if you supply a feature some people will abuse it (as in "abusing hints"), rather than using it with the caution it merits.
> I will leave this futile discussion now
Before you go, could you tell us what you would have to do as a DBA to ensure that Postgres could calculate the correct selectivity on something as simple as:
date_delivered > date_placed + 72 hours ;
-- Regards Jonathan Lewis http://jonathanlewis.wordpress.comReceived on Sat Feb 05 2011 - 08:21:40 CST