From owner-freebsd-isp Sun Nov 24 00:18:07 1996 Return-Path: owner-isp Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id AAA03358 for isp-outgoing; Sun, 24 Nov 1996 00:18:07 -0800 (PST) Received: from news.interworld.net (news.interworld.net [206.124.224.6]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id AAA03352 for ; Sun, 24 Nov 1996 00:17:58 -0800 (PST) Received: (from pete@localhost) by news.interworld.net (8.7.5/8.7.3) id AAA08700; Sun, 24 Nov 1996 00:17:56 -0800 (PST) To: freebsd-isp@freebsd.org Path: news.interworld.net!not-for-mail From: pete@news.interworld.net (Peter Carah) Newsgroups: freebsd.isp Subject: Re: Decision in Router Purchase Date: 24 Nov 1996 00:17:54 -0800 Organization: InterWorld Communications Lines: 43 Distribution: fbsd Message-ID: <5790bi$8fj@news.interworld.net> References: <199611141724.LAA25419@brasil.moneng.mei.com> Sender: owner-isp@freebsd.org X-Loop: FreeBSD.org Precedence: bulk In article <199611141724.LAA25419@brasil.moneng.mei.com>, Joe Greco wrote: > >> On Wed, 13 Nov 1996, Veggy Vinny wrote: ....... >> Go with the cisco! There is something just a bit off with freebsd's >> tcp/ip. I have a subgroup of users who get stalls, if my freebsd's are >> not the other side of my cisco from them. For instance, if they were to >> pull headers from a new server on the same subnet, the news server being >> freebsd, it would stop.. Same with web pages. > >One of my clients is an ISP with well over a thousand (local) lines, >they have not reported any such problems to me... the news servers >are FreeBSD. >Just a data point, For another data point, we had a somewhat similar problem that turned out to have been caused by a bad ethernet hub. I'd suspect that a lot of tcp/ip problems are really bad ethernet configurations (too many cascaded hubs, bad hubs, bad wire, etc etc...) All of our hosts are either freebsd or macs (all using the newer open-transport except for one). BTW, it was a BEAR to find... Main symptom was out-of-sequence tcp retry packets, there was not any obvious excess of collisions either detected or not ("late collision" is often one not detected by the hub. I don't know if any of the freebsd drivers give that report.). My only problems with Bay AN routers is that they won't take enough ram to do a full 2-carrier peering anymore. I suspect that a 2501 won't either... Doesn't help that the backbone carriers are bickering over direct-peering with each other and that 2 of the biggest now refuse to transit :-( *THAT* wasn't anywhere in the "vision" of the internet... (fixes things up so that our cooperative backup arrangement with another similar-sized isp (2 T1's vs 1 T1 and a 10mb connection) doesn't work because one of the T1 carriers won't accept our bgp routes to the other isp. so much for backup :-( -- Pete