From owner-freebsd-hackers Sun Jun 21 12:21:55 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id MAA18275 for freebsd-hackers-outgoing; Sun, 21 Jun 1998 12:21:55 -0700 (PDT) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from xena.mindspring.com (xena.mindspring.com [207.69.142.2]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id MAA18269 for ; Sun, 21 Jun 1998 12:21:53 -0700 (PDT) (envelope-from rsanders@xena.mindspring.com) Received: (from rsanders@localhost) by xena.mindspring.com (8.8.5/8.8.5) id PAA24842; Sun, 21 Jun 1998 15:21:54 -0400 (EDT) From: Robert Sanders To: hackers@FreeBSD.ORG Subject: Re: TweakDUN References: <199806200644.XAA24111@stennis.ca.sandia.gov> Mime-Version: 1.0 (generated by tm-edit 7.108) Content-Type: text/plain; charset=US-ASCII Date: 21 Jun 1998 15:21:54 -0400 In-Reply-To: bmah@CA.Sandia.GOV's message of "Fri, 19 Jun 1998 23:44:51 -0700" Message-ID: Lines: 25 X-Mailer: Gnus v5.6.4/XEmacs 20.3(beta19) - "Kiev" Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG bmah@CA.Sandia.GOV (Bruce A. Mah) writes: > I don't use dialups very often nowadays, but I dimly remember trying to > negotiate a *smaller* MTU on a downlink, in order to try to get better > interactive performance (mumble mumble, use IP TOS bits and smarter queueing, Reordering packets doesn't help on a dialup link if you have a bulk transfer going with 1500 byte packets. Even if you put "interactive" priority packets at the head of the queue, you may already have a 1500 byte packet in progress. Worse, most modems have significant local buffering to accomodate MNP/v.42bis compression so there may be even *more* than one packet's worth of delay depending on exactly how close to the wire the queuing algorithm lives. Some people have proposed "fragmenting" the packets at layer 2 (or 2.5, whatever PPP is) with MP so that interactive packets can be inserted into the middle, but I can't say off the top of my head whether that's going to be any more efficient than just using IP fragmentation and/or a small MSS. Somebody else already pointed out that VJ header compression significantly reduces packet overhead. On the other hand, with my network architect's hat on I don't like the idea of tripling the rate of packets per sec through already busy major exchanges and core routers. regards, -- Robert To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message