Date: Fri, 2 Feb 2001 08:46:31 +0800 From: "David Xu" <davidx@viasoft.com.cn> To: "Tony Finch" <dot@dotat.at>, "Kris Kennaway" <kris@obsecurity.org> Cc: "bsddiy" <bsddiy@163.net>, <freebsd-net@FreeBSD.ORG> Subject: Re: sendfile() Message-ID: <001d01c08cb1$9c445d80$6201a8c0@William> References: <1217774688.20010201133139@163.net> <20010201023825.A71975@xor.obsecurity.org> <20010201180010.Q70673@hand.dotat.at>
next in thread | previous in thread | raw e-mail | index | archive | help
----- Original Message ----- From: "Tony Finch" <dot@dotat.at> To: "Kris Kennaway" <kris@obsecurity.org> Cc: "bsddiy" <bsddiy@163.net>; <freebsd-net@FreeBSD.ORG> Sent: Friday, February 02, 2001 2:00 AM Subject: Re: sendfile() > Kris Kennaway <kris@obsecurity.org> wrote: > Linus's argument against the HP/UX and BSD sendfile() is that, for > example, in the presence of pipelined HTTP requests it isn't possible > to avoid poor packet boundaries between responses. TCP_NOPUSH does > help to solve this problem (it was designed to hack around problems > with the interaction between mbuf sizes and sosend and tcp_output) but > it introduces a new problem: at the end of a chain of HTTP responses > you want the last segment to go out quickly rather than waiting for > more data to fill the segment. For this reason turning off TCP_CORK > pushes out queued data, but this isn't the case for TCP_NOPUSH. > if TCP_NOPUSH can simulate TCP_CORK when TCP_NOPUSH is turned off and last data segment is sent, I'll be satisfied. but as I know, it seems TCP_NOPUSH is mainly used for TTCP, right? the idea behind TCP_CORK is it buffers any small data segment user program sending until these segments full fills a max TCP packet, then the packet is sent, web servers always send many very small HTTP headers, cause lots of small packets sent out, TCP_CORK can increase network performance. Regards, David Xu N'rzǧvf&j:+v "ryy b^nrazg
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?001d01c08cb1$9c445d80$6201a8c0>