Date: Wed, 6 Mar 1996 15:10:52 -0800 (PST)
From: Mark Crispin <MRC@Panda.COM>
Sender: Mark Crispin <mrc@Ikkoku-Kan.Panda.COM>
Subject: Re: re PIPELINING
To: drums@cs.utk.edu
In-Reply-To: <19960306230035.15369.qmail@koobera.math.uic.edu>
Message-Id: <MailManager.826153852.7721.mrc@Ikkoku-Kan.Panda.COM>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII

I propose that we all ignore Bernstein until he listens to contrary opinions
without airily dismissing them as "Bullshit" followed by unsubstantiated
claims ("if you had even glaced at...", "it scales just fine", etc.).  His
current means of discussion is not condusive towards achieving understanding,
much less arriving at a solution.

On 6 Mar 1996 23:00:35 GMT, D. J. Bernstein wrote:
> > Bernstein's algorithm is intended to improve real-time latency with
> > sendmail's
> > type of queueing strategy.
>
> Bullshit. If you had even glanced at qmail you would be aware that its
> queueing strategy is very different from sendmail's.
>
> In particular, qmail does not ``delay the easy cases because of ... the
> hard cases.'' Messages are handled independently. The only reason qmail
> ever delays is to obey the concurrency limit imposed by the sysadmin.
>
> > Nor does it scale;
>
> Bullshit. It scales just fine.
>
> > I recommend that we make the observation that using a single-threaded FIFO
> > queue in a mailer leads to performance problems, and move on.
>
> ``FIFO queue'' is redundant. ``Single-threaded queue'' is meaningless.
> And it's counterproductive to call a particular level of performance a
> ``problem'' if you don't suggest a way to do better.
>
> I (obviously) have no objection to the existing ``multiple concurrent
> outgoing mail transactions'' paragraph.
>
> ---Dan