Return-Path: Delivered-To: new-httpd-archive@hyperreal.org Received: (qmail 6698 invoked by uid 6000); 6 Jan 1998 03:44:21 -0000 Received: (qmail 6692 invoked from network); 6 Jan 1998 03:44:20 -0000 Received: from valis.worldgate.com (marcs@198.161.84.2) by taz.hyperreal.org with SMTP; 6 Jan 1998 03:44:20 -0000 Received: from localhost (marcs@localhost) by valis.worldgate.com (8.8.7/8.8.7) with SMTP id UAA11171 for ; Mon, 5 Jan 1998 20:44:14 -0700 (MST) Date: Mon, 5 Jan 1998 20:44:14 -0700 (MST) From: Marc Slemko To: Apache list Subject: Re: setsockopt in child_main In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: new-httpd-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org On Mon, 5 Jan 1998, Marc Slemko wrote: > SO_DEBUG, SO_DONTROUTE, SO_KEEPALIVE, SO_LINGER, SO_OOBINLINE, SO_RCVBUF > and SO_SNDBUF are inherited; if not, your system is broken. Oh, cf. R. Stevens "Unix Network Programming", volume one, second edition, page... erm.... 183 for details. > > Some of them have to be, since after the connection is established it can > be too late to set them; eg. buffer sizes are related to the window scale > option which is negotiated before accept() returns. > > On Mon, 5 Jan 1998, Igor Tatarinov wrote: > > > Correct me if I am wrong but currently socket options (KEEPALIVE and > > SND_BUF, SO_LINGER) are only set for the listening sockets (in make_sock). > > > > If all these options were inherited on accept, everything would be fine but > > TCP_NODELAY seems to be an exception (disable_ nagle is also called in > > child_main). So it is not inherited? (I checked it isn't; throughput > > degarades significantly without it). > > > > I suspect all others are not inherited either so they have to be set again > > (actually the listening sockets do not probably require a large buffer). I > > may be wrong though. > > > > any ideas? > > > > thanks, > > igor > > >