From: <gnuoytr@rcn.com>
Subject: Re: Useless sort by
To: pgsql-performance@postgresql.org
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-Id: <20100923095116.ALF56984@ms14.lnh.mail.rcn.net>
Date: Thu, 23 Sep 2010 09:51:16 -0400 (EDT)

I can't tell if you meant for this to be insulting or my reading it that way is wrong, but it certainly wasn't put in a helpful tone.  Let me summarize for you.  You've been told that putting ORDER BY into a view is a generally poor idea anyway, that it's better to find ways avoid this class of concern altogether.  There are significant non-obvious technical challenges behind actually implementing the behavior you'd like to see; the concerns raised by Tom and Maciek make your idea impractical even if it were desired.  And for every person like yourself who'd see the benefit you're looking for, there are far more that would find a change in this area a major problem.  The concerns around breakage due to assumed but not required aspects of the relational model are the ones the users of the software will be confused by, not the developers of it.  You have the classification wrong; the feedback you've gotten here is from the developers being user oriented, not theory oriented or 
 c!
ode oriented.
-- 
Greg Smith, 2ndQuadrant US greg@2ndQuadrant.com Baltimore, MD
PostgreSQL Training, Services and Support  www.2ndQuadrant.us
Author, "PostgreSQL 9.0 High Performance"    Pre-ordering at:
https://www.packtpub.com/postgresql-9-0-high-performance/book


Not insulting, just amused bemusement.  PG portrays itself as the best OS database, which it may well be.  But it does so by stressing the row-by-agonizing-row approach to data.  In other words, as just a record paradigm filestore for COBOL/java/C coders.  I was expecting more Relational oomph.  As Dr. Codd says:  "A Relational Model of Data for Large Shared Data Banks".  Less code, more data.

robert