[development] Do not let postgresql hold back great patches

Chris Johnson cxjohnson at gmail.com
Mon Dec 3 15:51:18 UTC 2007


Beware the premature optimization.

Saying that one way to avoid these problems is to use lowest common
denominator SQL sounds awful on the surface, because -- horrors! -- we
can't use all those great, unique to one database only, functions we
optimize performance.

But in reality, very few of the database-unique SQL syntax is really
highly optimized behavior (e.g. count(*) on a MySQL MyISAM table).
Most of them are just "different" ways of doing things than standard
or than other databases.

concat() is not faster than standard SQL ||, for instance.  ;-)

We "waste" far more database performance by poorly designed tables,
poorly designed indexes and poorly designed algorithms than we ever
gain in MySQL unique features, I'll bet.

Let's be sure we focus our optimization effort in places it really
helps, and not let it get in the way of making the software more
portable and useable.

I now surrender the soapbox.


More information about the development mailing list