Date: Sun, 17 Mar 1996 21:18:02 -0800 (PST)
From: Mark Crispin <MRC@Panda.COM>
Sender: Mark Crispin <mrc@Ikkoku-Kan.Panda.COM>
Subject: re: multiple RCPTs, again
To: drums@cs.utk.edu
In-Reply-To: <19960318040410.20972.qmail@koobera.math.uic.edu>
Message-Id: <MailManager.827126282.3608.mrc@Ikkoku-Kan.Panda.COM>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII

I guess this is all anyone needs to conclude that any sort of technical
discussion with Bernstein is useless.

On 18 Mar 1996 04:04:10 GMT, D. J. Bernstein wrote:
>
> Crispin admits that multiple RCPTs aren't always a good idea. (He's
> still wrong about how _often_ they're a good idea; see below.)
>
> Press him harder, and he might even admit that the situations where
> it's a win can be discovered, on the fly, by the MTA.
>
> What does this say about the wisdom of _requiring_ multiple RCPTs?
>
> > This is the fallacy in the argument behind using multiple SMTP streams to
the
> > same site for the same message.  It assumes a certain SMTP server behavior
> > (slow RCPT response due to DNS validation) which, while common, is not
> > universal.
>
> Wrong once again.
>
> You get bitten by latency without that behavior---even on slow links.
>
> An example from Brian Edmonds: ``Aggregation [i.e., multiple RCPTs] can
> be a real pain over slow links. I run a few mailing lists from my home
> machine on the end of a 14.4 link, and due to my setup, everything goes
> out through an upstream mail hub. Sendmail aggregates all addresses into
> one SMTP connection and you can watch the RX/TX lights flick back and
> forth as it slowly grinds through all 200-300 addresses in lockstep.''
>
> Did you forget that slow links usually have high latency, Mark?
>
> ---Dan