From owner-freebsd-net Mon Jul 5 9:33:29 1999 Delivered-To: freebsd-net@freebsd.org Received: from vortex.greycat.com (vortex.greycat.com [207.173.133.4]) by hub.freebsd.org (Postfix) with SMTP id C9F8514F8F for ; Mon, 5 Jul 1999 09:33:08 -0700 (PDT) (envelope-from dann@greycat.com) Received: (qmail 6701 invoked from network); 5 Jul 1999 16:33:06 -0000 Received: from bigphred.greycat.com (HELO greycat.com) (207.173.133.2) by vortex.greycat.com with SMTP; 5 Jul 1999 16:33:06 -0000 Message-ID: <3780DE74.2DD4F526@greycat.com> Date: Mon, 05 Jul 1999 09:33:56 -0700 From: Dann Lunsford Organization: You're kidding, right? X-Mailer: Mozilla 4.5 [en] (X11; U; FreeBSD 3.2-STABLE i386) X-Accept-Language: en MIME-Version: 1.0 To: Net@freebsd.org Subject: mbuf-less IP? Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Just a question out of pure curiosity. I've been studying the Stevens set ("TCP/IP Illustrated", all three vols) and came across a comment that Van Jacobson had an experimental IP stack that didn't use the mbuf structures. The reasoning was that the current implementation was designed when memory was much more constrained, CPU speed was lower, and networks weren't as fast or fat. Reportedly, Jacobson found siginificant performance improvements and better resource utilization over current, mbuf based, implementations. So... Has anybody looked into this idea? I realize it would entail *MAJOR* rewrites and redesigns, and would not be undertaken lightly, but perhaps it would be worth considering for the future. The current design smacks of "It's always been done that way"; whenever I hear that phrase, a chill goes up my spine :-). Anyway, that's it. Just tossing it into the ring; please don't throw knives back :-). Dann Lunsford dann@greycat.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message