Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!apple!usc!elroy.jpl.nasa.gov!sdd.hp.com!hplabs!pyramid!infmx!barrym
From: barrym@informix.com (Barry Mednick)
Newsgroups: comp.databases
Subject: Re: how bad is oracle? (really: query optimizers)
Message-ID: <1991Feb19.204354.16211@informix.com>
Date: 19 Feb 91 20:43:54 GMT
References: <1991Feb3.182558.14411@oracle.com> <70683@microsoft.UUCP> <13512@blia.sharebase.com>
Sender: Barry Mednick
Organization: Informix Software, Inc.
Lines: 12

In article <13512@blia.sharebase.com> mike@woodstock.UUCP (Mike Ubell) writes:
>
>True enough, but: A good DBMS should beable to optimize at compile time
>and find the best query plan without help from the user and then store
>the plan so that time critical applications have acceptible performance.
Without help from the user, yes.  At compile time, perhaps not.
Consider what could change between compile time and run time:
  Indexes can be dropped and created
  The sizes of tables can change
All the factors used to determine a good query plan may be different
when the program is run.  Wouldn't it make more sense to optimize
at run time?