Re: [tcpm] Is this a problem?

Ted Faber <faber@ISI.EDU> Wed, 21 November 2007 17:34 UTC

Return-path: <>
Received: from [] ( by with esmtp (Exim 4.43) id 1IutT1-00018T-SU; Wed, 21 Nov 2007 12:34:27 -0500
Received: from tcpm by with local (Exim 4.43) id 1IutT0-00018C-4Z for; Wed, 21 Nov 2007 12:34:26 -0500
Received: from [] ( by with esmtp (Exim 4.43) id 1IutSz-00016y-P7 for; Wed, 21 Nov 2007 12:34:25 -0500
Received: from ([]) by with esmtp (Exim 4.43) id 1IutSv-0004sS-Ur for; Wed, 21 Nov 2007 12:34:25 -0500
Received: from ( []) by (8.13.8/8.13.8) with ESMTP id lALHWrn7020439 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NOT); Wed, 21 Nov 2007 09:32:54 -0800 (PST)
Received: (from faber@localhost) by (8.14.2/8.14.2/Submit) id lALHWnd2014435; Wed, 21 Nov 2007 09:32:49 -0800 (PST) (envelope-from faber)
Date: Wed, 21 Nov 2007 09:32:49 -0800
From: Ted Faber <faber@ISI.EDU>
To: Lloyd Wood <>
Subject: Re: [tcpm] Is this a problem?
Message-ID: <>
References: <> <> <>
Mime-Version: 1.0
In-Reply-To: <>
User-Agent: Mutt/
X-ISI-4-43-8-MailScanner: Found to be clean
X-Spam-Score: 0.0 (/)
X-Scan-Signature: 8b431ad66d60be2d47c7bfeb879db82c
X-Mailman-Version: 2.1.5
Precedence: list
List-Id: TCP Maintenance and Minor Extensions Working Group <>
List-Unsubscribe: <>, <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
Content-Type: multipart/mixed; boundary="===============0954148168=="

Speaking as myself:

On Wed, Nov 21, 2007 at 05:46:12AM +0000, Lloyd Wood wrote:
> At what point does a resource stop being shared and start being local
> (or vice versa)? It seems a bit fuzzy to me.

So, one can make these waters pretty deep.  Communication resources are
shared at some level just because they're being used to swap data
between hosts.  I wouldn't be using a TCP buffer if I wasn't talking to
anyone else, and using it to talk to you precludes me from using it to
talk to Mark.  They are all shared resources.  However for scalability
and convenience, they're all locally administered.  The power to destroy
a thing ...

TCP standards (by and large) recognize that split and are pretty quiet
on how routers and end hosts do the administration.   That's a feature,
IMHO.  TCP standards don't tell hardware designers/application writers
how to make the optimal use of their resources; the standards keep them
from breaking the protocol.  Mostly.

Again, as myself, I prefer to let the TCP standards proscribe resource
allocation systems that breaks the protocol (not applications) rather
than require resource allocation systems that optimize some use of the
protocol.  Even extremely prevalent uses.

Ted Faber           PGP:
Unexpected attachment on this mail? See
tcpm mailing list