From owner-freebsd-current Tue Mar 14 12:56:35 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id MAA00424 for current-outgoing; Tue, 14 Mar 1995 12:56:35 -0800 Received: from pelican.com (pelican.com [134.24.4.62]) by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id MAA00418 for ; Tue, 14 Mar 1995 12:56:31 -0800 Received: by pelican.com (Smail3.1.28.1 #5) id m0roa0a-000K3UC; Tue, 14 Mar 95 09:03 WET Message-Id: From: pete@pelican.com (Pete Carah) Subject: Re: bad outgoing serial coms To: jhs@regent.e-technik.tu-muenchen.de (Julian Howard Stacey) Date: Tue, 14 Mar 1995 09:03:48 -0800 (PST) Cc: current@FreeBSD.org In-Reply-To: <199503132234.XAA15856@vector.enet> from "Julian Howard Stacey" at Mar 13, 95 11:34:35 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1654 Sender: current-owner@FreeBSD.org Precedence: bulk Julian Howard Stacey writes: > > Make TTYHOG big > will try that, > thanks, > will also check the flow control stuff, though I've not changed anything there in ages Right. And Bruce swears up & down that ttyhog isn't used in slip or ppp, but there really is only one place that message comes from... Actually it *does* apply to the new 'process' version of ppp (/usr/sbin/ppp as opposed to /usr/sbin/pppd). Are you using that? My friend here had to up ttyhog from his usual 10k to 12k lately to quiet that message; this could be related to the big flurry of commit messages that recently came through (they are short; the problem with Taylor uucp mostly comes up on file boundaries, and we are now using a 28.8k modem-pair for our uucp link). G protocol doesn't window message packets but i does. We are both using smail 3 instead of sendmail; with this the problem for mail could be alleviated somewhat by using rsmtp, but that doesn't change the base problem, which is that ttyhog needs to be somewhat dynamic to handle large-window protocols (Taylor I/i protocol at 16k window; kermit/big can get even bigger; the new user-mode ppp (which doesn't use the ppp line discipline module) would need ttyhog at probably twice the tcp window size plus some for the ppp encapsulation sequences (tcp window applies per connection, so if one has external fast feeds and does ftp, ttyhog might have to be 3 or 4 times the tcp window. I'd think that normally that much shouldn't be needed). A side note, discovered while looking through the sources for this: Nothing appears to ever fill in tty_tb's row in linesw; has this ever been tested? -- Pete