Date: Thu, 6 Jul 2006 08:51:41 -0400
From: Andrew Sullivan <ajs@crankycanuck.ca>
To: pgsql-sql@postgresql.org
Subject: Re: Alternative to serial primary key
Message-ID: <20060706125141.GE26210@phlogiston.dyndns.org>
References: <12b7ac1e0607060316ub7b5e21i5e8e3bef2504e02c@mail.gmail.com>
	<44ACEA1C.9030804@logix-tt.com>
	<12b7ac1e0607060406r16d03f74mfa303eec9ab4821f@mail.gmail.com>
	<44ACEFBF.7030903@logix-tt.com>
	<20060706111223.GC26210@phlogiston.dyndns.org>
	<12b7ac1e0607060432i5fd3aa4ap16af10b9efabb917@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <12b7ac1e0607060432i5fd3aa4ap16af10b9efabb917@mail.gmail.com>
User-Agent: Mutt/1.5.9i

On Thu, Jul 06, 2006 at 11:32:36PM +1200, David Clarke wrote:
> 
> Yes, that occurred to me as well. Frankly I believe the md5 collision
> generation is more of a practical issue for crypto where for my
> purposes the potential for two residential street addresses to
> generate the same md5 hash value is effectively zero. And the md5
> function is a builtin which I would hope is faster than anything I
> could write in pgsql. Could be wrong, I have been before.

You could, of course, just put a unique index on it and accept that,
in the event of collision, you'll have to cope with the error.  It's
probably an acceptable cheat, as you're right that the collision risk
is pretty small.

A

-- 
Andrew Sullivan  | ajs@crankycanuck.ca
If they don't do anything, we don't need their acronym.
		--Josh Hamilton, on the US FEMA