Date: Wed, 7 Sep 2005 17:19:12 +0000
From: nathan wagner <nw@hydaspes.if.org>
To: pgsql-hackers@postgresql.org
Subject: Re: uuid type for postgres
Message-ID: <20050907171912.GA6427@granicus.if.org>
References: <20050906135057.GA24136@granicus.if.org>
	<200509061840.28068.josh@agliodbs.com>
	<20050907023606.GA2181@granicus.if.org>
	<200509070948.08307.josh@agliodbs.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <200509070948.08307.josh@agliodbs.com>
User-Agent: Mutt/1.4.2.1i

On Wed, Sep 07, 2005 at 09:48:08AM -0700, josh@agliodbs.com wrote:

> That's good, it gives users options.  And I can see why you don't want to 
> re-create the functionality in PG code, it's probably pretty large.

It would also be something else that would have to be maintained and
debugged.  Which i would be starting from scratch.  The OSSP
UUID library is at version 1.3.  From this I infer that he has spent
some time maintaining and debugging it.  The less duplication of
code and effort the better.

There may be a problem with using the library as a core type though,
due to windows portability concerns.  That said, it would probably
be easier to port the underlying library than to start from scratch.

> Well, IMHO, if the unguessability of UUIDs is an important part of your spec, 
> you have some design problems.

I was thinking of cryptographic applications.  Then, to use an
example from another domain, initial TCP sequence numbers should
be random (i.e. unguessable).  A problem with TCP perhaps.  It's
been a while since i've read over my copy of _Applied Cryptography_,
but I seem to recall that unguessable numbers were sometimes useful.

-- 
Nathan Wagner