From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Hemminger Subject: Re: [PATCH] (1/3) tcp - choose congestion algorithm at initialization Date: Tue, 28 Sep 2004 09:08:54 -0700 Sender: netdev-bounce@oss.sgi.com Message-ID: <1096387734.21799.1.camel@localhost.localdomain> References: <20040927111834.48c7baab@zqx3.pdx.osdl.net> <20040927225858.3a1665f1.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: Herbert Xu , davem@redhat.com, netdev@oss.sgi.com Return-path: To: "David S. Miller" In-Reply-To: <20040927225858.3a1665f1.davem@davemloft.net> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Mon, 2004-09-27 at 22:58 -0700, David S. Miller wrote: > On Tue, 28 Sep 2004 08:07:36 +1000 > Herbert Xu wrote: > > > Stephen Hemminger wrote: > > > The choice of congestion algorithm needs to be made when connection > > > is setup to avoid problems when the sysctl values change later and the > > > necessary data hasn't been collected. > > > > Could this be chosen by a setsockopt as well? > > If we export such things with a user visible > interface, that makes it harder to rip out the > congestion control algorithm later. Such an > action is very likely so... The current set of algorithm's is because we haven't found the right one. If it turns out there are multiple algortihm's that make sense to support long term, then the choice should be done in as route hints not as part of the user API.