From owner-freebsd-hackers Sat May 31 13:29:49 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id NAA10979 for hackers-outgoing; Sat, 31 May 1997 13:29:49 -0700 (PDT) Received: from time.cdrom.com (root@time.cdrom.com [204.216.27.226]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id NAA10974 for ; Sat, 31 May 1997 13:29:46 -0700 (PDT) Received: from time.cdrom.com (jkh@localhost [127.0.0.1]) by time.cdrom.com (8.8.5/8.6.9) with ESMTP id NAA09909 for ; Sat, 31 May 1997 13:30:13 -0700 (PDT) To: hackers@freebsd.org Subject: LINT and GENERIC - between a rock and a generic place. Date: Sat, 31 May 1997 13:30:12 -0700 Message-ID: <9876.865110612@time.cdrom.com> From: "Jordan K. Hubbard" Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk More and more people are trying to use GENERIC as a template for their own kernels and they're losing, of course, because generic sets many limits (like max children or open files) too low. Unfortunately, every time this issue has come up in the past it's also followed a rather set course, in 3 distinct stages: 1. Someone starts off the discussion with "Well, why don't we just make up some canned NEWS_SERVER, DESKTOP, WEB_SERVER and so on config files? We can put comments at the beginning of each about how the user should customize that file to fit certain situations and Bob's your bloody uncle, we're there!" Things would probably progress well from there if it weren't for the stage-II participants in this discussion who then come up and say: 2. "Hey, that's gross. What we really need here is a config file metaformat which encodes all the information in LINT in such a way that menus and choiceboxes and such can be build dynamically to configure all the items there, and if you want canned configs then then we just bundle them with the config tool as starter-settings and we don't have to clutter up /sys/i386/conf. Say, has anyone looked at Linux's ``make xconfig?'' 3. The discussion now veers into a debate on the merits/evils of Linux's kernel configurator, how config is just totally broken anyway and we really need to get rid of it and its lousy config files altogether, so on and so forth. The first group slinks away, ashamed that they could have even proposed such a low-tech solution as putting more sample files into /sys/i386/conf. Meanwhile, of course, the users continue to use GENERIC (or worse, LINT) as their only available guides and they continue to walk off cliffs, year after year. How shall we conduct the debate this time? Same old same old, or something genuinely productive? :-) Jordan