To:
Patrik Fältström <paf@cisco.com>
cc:
Edward Lewis <lewis@tislabs.com>, "'ietf-provreg@cafax.se'" <ietf-provreg@cafax.se>, Allison Mankin <mankin@isi.edu>, Scott Bradner <sob@harvard.edu>, Ned Freed <ned.freed@mrochek.com>
From:
Allison Mankin <mankin@isi.edu>
Date:
Fri, 11 Jan 2002 08:14:05 -0500
In-reply-to:
Your message of Thu, 10 Jan 2002 22:02:05 -0800. <330757.1010700125@localhost>
Reply-To:
mankin@isi.edu
Sender:
owner-ietf-provreg@cafax.se
Subject:
Re: Requirements Document Update
Patrik, I would only add to what you and Scott wrote that: > _somewhere_ in the protocol stack there has to be congestion control It is not only a large undertaking to put congestion control in the application itself, but it may greatly distort the application protocol if it has the requirement to make the network timing measurements needed for congestion control (taken care of by using TCP or SCTP). We asked for the wording in the provreg document to encourage provreg not to choose UDP as the transport and then be forced to provide application congestion control as well. Allison Allison