From: Christopher Browne <cbbrowne@acm.org>
Subject: Re: Joel's Performance Issues WAS : Opteron vs Xeon
Date: Sat, 23 Apr 2005 14:27:12 -0400
Organization: cbbrowne Computing Inc
Lines: 29
Message-ID: <m3sm1hs60v.fsf@knuth.cbbrowne.com>
References: <20050423021107.GX58835@decibel.org>
	<001701c5480e$57e7fad0$797ba8c0@jfradkin>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4 (Jumbo Shrimp, linux)
Cancel-Lock: sha1:TzciKb9rv6zK1tANo1A8n6iSgHU=
To: pgsql-performance@postgresql.org

Centuries ago, Nostradamus foresaw when jfradkin@wazagua.com ("Joel Fradkin") would write:
> I am just testing the water so to speak, if it cant handle single
> user tests then multiple user tests are kind of a waste of time.

I would suggest that if multi-user functionality is needed, then
starting with single user tests is a similar waste of time.

There's good reason to look at it this way...  

It is all too common for people to try to start building things with
primitive functionality, and then try to "evolve" the system into what
they need.  It is possible for that to work, if the "base" covers
enough of the necessary functionality.

In practice, we have watched Windows evolve in such a fashion with
respect to multiuser support, and, in effect, it has never really
gotten it.  Microsoft started by hacking something on top of MS-DOS,
and by the time enough applications had enough dependancies on the way
that worked, it has essentially become impossible for them to migrate
properly to a multiuser model since applications are normally designed
with the myopic "this is MY computer!" model of the world.

You may not need _total_ functionality in the beginning, but,
particularly for multiuser support, which has deep implications for
applications, it needs to be there In The Beginning.
-- 
output = reverse("moc.liamg" "@" "enworbbc")
http://linuxdatabases.info/info/lisp.html
A CONS is an object which cares.  -- Bernie Greenberg.