Re: Surrogate Keys: an Implementation Issue

From: -CELKO- <jcelko212_at_earthlink.net>
Date: 22 Jul 2006 15:10:56 -0700
Message-ID: <1153606256.006641.26630_at_i42g2000cwa.googlegroups.com>


>> Replace programmer or program with user, and what you have stated is a decent goal. ..
.. the physical encoding of a pointer is internal to the dbms, while the DBA can specify how to physically arrange pointers into indexes, lists etc. Likewise, the DBA can specify what data to cluster together without dealing directly with the internal representations. <<

I will go one step beyond that. I want to write in a high level language and have no idea if the engine is using indexes, bit vectors, pointers or whatever. I want the engine to change those methods, perhaps without even telling me, when it finds better ones (see Ingres indexing methods). I do not want to specify anything at the physical level.

I also want to win the lottery :)

With no proof, I think that optimizing an RDBMS to be somewhere between playing Grand Master Chess and playing tournament level Go. The machines are getting to that level of Chess today; Go will take longer.  But it is coming. My Grandfather saw the Wright Brothers and the Lunar landing in his lifetime; and today the rate of the ratre of change is increasing. Received on Sun Jul 23 2006 - 00:10:56 CEST

Original text of this message