Re: Update identification and length

barns@cove.mitre.org Mon, 04 May 1992 14:21 UTC

Received: from nri.nri.reston.va.us by ietf.NRI.Reston.VA.US id aa02389; 4 May 92 10:21 EDT
Received: from nri.reston.va.us by NRI.Reston.VA.US id aa24764; 4 May 92 10:26 EDT
Received: from PARK-STREET.BBN.COM by NRI.Reston.VA.US id aa24760; 4 May 92 10:26 EDT
Received: from park-street by PARK-STREET.bbn.COM id aa16521; 4 May 92 10:09 EDT
Received: from BBN.COM by PARK-STREET.BBN.COM id aa16517; 4 May 92 10:08 EDT
Received: from gateway.mitre.org by BBN.COM id aa24300; 4 May 92 10:03 EDT
Return-Path: <barns@cove.mitre.org>
Received: from [128.29.31.100] by gateway.mitre.org (5.61/SMI-2.2) id AA20080; Mon, 4 May 92 10:02:51 -0400
Received: by cove.mitre.org (4.1/SMI-4.1) id AA05474; Mon, 4 May 92 09:58:38 EDT
Message-Id: <9205041358.AA05474@cove.mitre.org>
To: Robert Woody Woodburn <woody@sparta.com>
Cc: idpr-wg@bbn.com, barns@cove.mitre.org
Subject: Re: Update identification and length
In-Reply-To: Your message of "Mon, 04 May 92 09:30:56 EDT." <9205041330.AA01284@crusher>
Date: Mon, 04 May 92 09:58:30 -0400
From: barns@cove.mitre.org

I don't know if I'm an opponent of Routing Over TCP, but I seem to
remember Dennis Ferguson mumbling something to the effect that the
VJ TCPs back off so aggressively in response to congestion that you
get some undesirable delay or timeout effects if backoff kicks in,
so you either don't want to use links as slow as 56kb or should use
priority mechanisms to protect the routing traffic.  This was something
he observed somewhere real, but I don't recall specifics.

Personally I think priority to protect the routing traffic is a rather
reasonable idea, and I also have this attitude that TCP backoffs per
accepted VJ religion may be more aggressive than is good for us, but
this kind of statement alarms the converts to binary exponential backoff.
However I note that the theory of these things does not say that you have
to have K**N with K=2 to recover, but rather K>1.  But I digress...

/Bill Barns