IETF Logo Mail Archive

Re: internet draft on suggested mod to the Nagle algorithm

Kacheong Poon <Kacheong.Poon@Eng.Sun.COM> Sat, 13 February 1999 01:25 UTC

X-Authentication-Warning: assateague-fi.lerc.nasa.gov: listserv set sender to owner-tcp-impl@lerc.nasa.gov using -f
Date: Fri, 12 Feb 1999 17:25:46 -0800
From: Kacheong Poon <Kacheong.Poon@Eng.Sun.COM>
Reply-To: Kacheong Poon <Kacheong.Poon@Eng.Sun.COM>
Subject: Re: internet draft on suggested mod to the Nagle algorithm
To: tcp-impl@lerc.nasa.gov
In-Reply-To: "Your message with ID" <36C39FDB.9EECFAA3@cup.hp.com>
Message-ID: <Roam.SIMCSD.2.0.4.918869146.24550.kcpoon@jurassic>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET="US-ASCII"
Sender: owner-tcp-impl@lerc.nasa.gov
Precedence: bulk
Status: RO
Content-Length: 761
Lines: 19

> Would a PTMU smaller than the MSS interfere with the receiver's ability
> to tell a full-sized segment from a small one? Especially in a
> unidirectional case, or an assymetric route case.

Yup, this can be a problem.  This is the delayed ack problem described in 
draft-ietf-tcpimpl-cong-control-03.txt, SMSS <> RMSS.  There are some
solutions proposed, though they are not "standardised."  But I guess this
can be worked around.  And there are stacks which already implement solutions
to handle the delayed ack problem.

> Well, it is my belief that "today" an ACK costs just as many CPU cycles
> as a data segment.

Sometimes the cost can be justified...  I think it is OK in this case to
avoid the extra delay.

							K. Poon.
							kcpoon@eng.sun.com

v2.23.0 | Report a Bug | By Email