From owner-freebsd-hackers Fri Nov 20 02:03:16 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id CAA12670 for freebsd-hackers-outgoing; Fri, 20 Nov 1998 02:03:16 -0800 (PST) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from mail.ruhrgebiet.individual.net (in-ruhr.ruhr.de [141.39.224.38]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id CAA12658 for ; Fri, 20 Nov 1998 02:03:09 -0800 (PST) (envelope-from bs@adimus.de) Received: (from admin@localhost) by mail.ruhrgebiet.individual.net (8.8.5-r-beta/8.8.5) with UUCP id KAA28389; Fri, 20 Nov 1998 10:41:22 +0100 (MET) Received: from mail by mx.adimus.de with local (Exim 1.92 #1) id 0zgmq6-0000A7-00; Fri, 20 Nov 1998 10:26:54 +0100 Received: from det.adimus.de(192.168.0.1) via SMTP by adimus.de, id smtpdIhQ590; Fri Nov 20 10:26:48 1998 Received: from bs by det.adimus.de with local (Exim 1.92 #1) id 0zgmpy-0000tr-00; Fri, 20 Nov 1998 10:26:46 +0100 To: Stephen McKay Cc: Bill/Carolyn Pechter , freebsd-hackers@FreeBSD.ORG Subject: Re: SysV Init References: <199811191220.HAA09760@shell.monmouth.com> <199811200546.PAA25250@nymph.dtir.qld.gov.au> Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit From: Benedikt Stockebrand Date: 20 Nov 1998 10:26:44 +0100 In-Reply-To: Stephen McKay's message of "Fri, 20 Nov 1998 15:46:02 +1000" Message-ID: Lines: 81 X-Mailer: Gnus v5.5/XEmacs 20.4 - "Emerald" Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Stephen McKay writes: > On Thursday, 19th November 1998, Bill/Carolyn Pechter wrote: > > >This is one area where SysV is superior. What's "superior"? Just additional functionality? Then it definitely *is* superior. Easier handling? Then it's a big step in the wrong direction. Which init is "superior" depends on your particular needs. And SysV init has been developed to cover someones needs. I know at least one job where a SysV init would be very handy: I'm currently running a firewall setup on FreeBSD. I use the securelevel kernel variable and chflags schg everything when I boot the machine. To do actual reconfiguration things on those machines I need to reboot them to reset the securelevel. I can't simply go into single user mode because those boxes don't have a console or anything connected. For maintenance I need to up an internal interface and start the sshd. With a SysV init I'd just use a "spare" runlevel for this. Now I use a custom rc file that checks for the existence of some "signal" files (like the old /fastboot thing) and then runs a different set of scripts to start the machine. But this requires to reboot the machine every time I want As a general rule of thumb I say that any machine that I need to handle primarily via remote administration has good use for a "remote administration" mode (or runlevel) in addition to the single user mode available with the BSD init. And there are probably other cases where a SysV init is useful. But in a "standard scenario" the SysV init will provide no useful functionality but instead cause unnecessary problems with its complexity. > Woah! Put me down as absolutely against this position. Put my face on > your dart board, if you must! Nah, we'll have you rewrite the FreeBSD rc scripts for SysV init :-) Actually, making a SysV init use standard BSD-style rc scripts isn't *that* much of a problem. And I strongly propose to keep the rc scripts BSD-style because they're usually easier to understand especially by newcomers. > >> I'm asking for a system where the legacy rc is there for those who want > ^^^^^^ > "Legacy" in this case is a nasty word. It's not legacy as far as I am > concerned. It is the current and best practice. I would be bitterly > disappointed if it was downgraded to a compatibility option on a SysV > style init system. Not even a compatibility option but a special inittab that makes it behave like a BSD init. > >My plan was two sysctl variables for current and past run state. > >This would avoid the need for a utmp change. > > > >kern.current_runlevel > >kern.prev_runlevel Is it necessary to mess around in the kernel for this? I'd try to keep runlevel information within the init instead. But I admit that I'm not comfortable enough with SysV init to know the conventions how it provides runlevel information to regular userland programs. If no such convention exists, why not use a Unix domain socket or such? > I'll take the lateral view on this and assume you mean compatibility with > existing practice. :-) I hope that any additions you make will not > cause any *requirement* to use run levels for any purpose whatsoever. > Optional use among consenting adults is hard to stop. \begin{aol}Me too\end{aol} So long, Ben -- Benedikt Stockebrand, Dipl. inf. Adimus Beratungsgesellschaft für System- und Netzwerkadministration mbH & Co KG System Administration & Design, Universitätsstr. 142, 44799 Bochum IT Security, Remote System Mgmt Tel. (02 34) 971 971 -2, Fax -9 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message