From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [patch 3/3] tcp: remove experimental variants from default list Date: Mon, 12 Feb 2007 12:47:34 -0800 (PST) Message-ID: <20070212.124734.07455258.davem@davemloft.net> References: <5640c7e00702121213p1430891el76e259aacd517c4a@mail.gmail.com> <20070212.123240.21597176.davem@davemloft.net> <20070212123713.1e8efb0d@freekitty> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: ian.mcdonald@jandi.co.nz, baruch@ev-en.org, netdev@vger.kernel.org To: shemminger@linux-foundation.org Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:41693 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S965336AbXBLUrg (ORCPT ); Mon, 12 Feb 2007 15:47:36 -0500 In-Reply-To: <20070212123713.1e8efb0d@freekitty> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org From: Stephen Hemminger Date: Mon, 12 Feb 2007 12:37:13 -0800 > My patches weren't reactionary. Going to pure old Reno is reactionary. > It was more looking at the state of the code on the flight back > and cleaning house. Others were/are reactionary. Ok. The only patch I have a real problem with is the DEFAULT_* removals, the choices are frankly arbitrary. Vegas is buggy, that's nice, why don't we simply fix the bugs in our implementation? Westwood is very conservative, frankly, and I therefore see no reason it cannot be offered as a default either. HTCP doesn't do anything earth shattering either. I think the whole suite of algorithms in that list are reasonable. And even re-reading your patch, you're messing with the DEFAULT_* setting for the case where the user selected TCP_CONG_ADVANCED. I think TCP_CONG_ADVANCED implies an intention by the user, and if he wants to choose one of those listed as a default why should we stop them? The distributions take the default we recommend, and that's all that matters for wide deployment. > I push the problem back in their court: "Why do you not have a process > that causes consensus?" IETF has done nothing to create any incentive > for long term cooperation. Yep, this is a good point. > Do I need to dig out the "Why Reno sucks" graphs? Hehe :-)