From owner-freebsd-arch Sun Dec 9 10:19:40 2001 Delivered-To: freebsd-arch@freebsd.org Received: from ns.yogotech.com (ns.yogotech.com [206.127.123.66]) by hub.freebsd.org (Postfix) with ESMTP id DFD5D37B417 for ; Sun, 9 Dec 2001 10:19:33 -0800 (PST) Received: from caddis.yogotech.com (caddis.yogotech.com [206.127.123.130]) by ns.yogotech.com (8.9.3/8.9.3) with ESMTP id LAA11979; Sun, 9 Dec 2001 11:19:10 -0700 (MST) (envelope-from nate@yogotech.com) Received: (from nate@localhost) by caddis.yogotech.com (8.11.6/8.11.6) id fB9IJ9183787; Sun, 9 Dec 2001 11:19:09 -0700 (MST) (envelope-from nate) From: Nate Williams MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <15379.43805.336137.177646@caddis.yogotech.com> Date: Sun, 9 Dec 2001 11:19:09 -0700 To: Matthew Dillon Cc: Jordan Hubbard , Bernd Walter , Garance A Drosihn , "Louis A. Mamakos" , Sheldon Hearn , Kirk McKusick , freebsd-arch@FreeBSD.ORG Subject: Re: Proposed auto-sizing patch to sysinstall (was Re: Using a larger block size on large filesystems) In-Reply-To: <200112090941.fB99fGV36341@apollo.backplane.com> References: <50925.1007888526@winston.freebsd.org> <200112090941.fB99fGV36341@apollo.backplane.com> X-Mailer: VM 6.96 under 21.1 (patch 14) "Cuyahoga Valley" XEmacs Lucid Reply-To: nate@yogotech.com (Nate Williams) Sender: owner-freebsd-arch@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Matthew Dillon writes: > Sigh. Look, the whole point of 'A'uto is to create a reasonable > setup for a layperson installing a system. It is not there to > cow-tow to a minimalist status-quo. It is there to give a layperson > a reasonable system that does not require him to screw around > with the infrastructure when he does reasonable things, like add an > account for themselves or install a bunch of ports or follow OUR > directions on how to retrieve, compile, and install system source. So far we agree. > That is what the option is there for and right now sysinstall doesn't > even come CLOSE to providing that. This is where we disagree. I just installed 4 systems in the last month, and all were installed using 'auto' partitions, and all are using defaults that are acceptable for single-user systems. > It creates partitions that are > too small. I disagree. They may not be optimal, but they are acceptable. > It creates relatively unsafe partitions - for example, > leaving /var/tmp on /var where /var itself is ALREADY too small for > a number of ports, including our printing mechanism and vmware. Completely disagreed. /var/tmp doesn't need to be any bigger *IF* you don't symlink /tmp into /var/tmp. (Which I still think is a *REALLY* *REALLY* *BAD* idea, but unfortunately I'm certain this will become the point to argue about, because I think this is the basis for most of your othe changes. :() > You > may be smart enough to ensure that your mail spool and /var/tmp > don't fill up and screw each other over (it can go in either direction), > but the layperson is NOT. Disagreed. For most single-user systems, or 'server-type' systems, the defaults we're using now aren't that bad. For shell servers used by ISP's, or other servers with specific purposes, the defaults aren't that great, but if you're building one of those, you shouldn't be using the defaults and you should be clueful enough to know how to build a big system. (There are tons of great resource to help you out on this, including books, mailing lists, etc....,) > You may be smart enough to know how to > manage /usr but the layperson is not. We can't solve every problem > in sysinstall but we can easily and trivially solve a number of them. > > My patch creates a far safer default partitioning for the layperson > and, you know what? I think it creates a far better default partitioning > for many FreeBSD developers and power users as well. It is certainly > far, far, FAR superior to what sysinstall does now. That is where we disagree. And, many of the other developers disagree that your scheme is 'superior' to the current defaults. Hence, the quandry. If so many people disagree, I think it implies that your opinions don't necessarily reflect the opinions of others who are *equally* qualified to judge your defaults. Therefore, either we're all completely clueless (my guess is that's your opinion of the matter), or that your defaults aren't complete, and need more than what you've done now to be a better solution than what we have now. Changing the defaults doesn't necessarily make things better, and in my opinion I think they will make things *more* difficult for many types of configuration. Having lots of small partitions is a step in the wrong direction for *many* installations, because when /var fills up, or /usr fills up, or /home fills up, there is almost invariably space in the other partitions that they can't use. Having '/' on one partition (appropriate sized, but as small as possible for fsck to be able to fix it quickly), having /var (so that / doesn't need to be written to decreasing the chances of having FS corruption, plus /var tends to fill up while / does), and having /usr is a good first start. Then, having the ability to create an memory-based /tmp is a good next step, since for most people /tmp *should* be files that don't stick around much, and enabling soft-updates on / doesn't work so well when disk space gets tight. Finally, having the ability to create a separate /home partition is the next logical step, since /home is for most folks the directory for logins. From here, we can get nuts, and start creating /var/tmp, and /usr/obj, and /usr/local, and all sorts of other partitions. But, I think we either need to stick with the current very simple (and *VERY* workable defaults), or provide a more flexible mechanism so that the user can get reasonable defaults without forcing them into an out-of-diskspace setup. Note, *MANY* of the people that use FreeBSD are installing it on 'older' hardware that want to try it out, or are sticking FreeBSD into places where there isn't much 'user' traffic on the box aside from email. These kinds of machines don't need /home, or /var/tmp. > > Now I've spent the time and effort to fix this, and I am getting rather > sick and tired of people trying to impose a minimalist view on > sysinstall's actions on the one hand, and a complex view on the other. Because sysinstall > I am sick and tired of people who complain that it doesn't reflect their > view of reality. Well, guess what? IT NEVER WILL! These views are > why sysinstall's auto-partitioning has essentially been broken for > years now. This is where we disagree. The only one who has been complaining about it being completely broken is you. Now, I'm not saying it's been optimal, but it hasn't been broken. > Hell, all of you have had YEARS to fix this and haven't lifted a finger. > You all have complained and argued a lot, but code? I haven't complained about sysinstall. I've complained about you're current forcing your opinion into something that isn't broken, thus making the 'Auto' feature much less useful for me and a number of folks I support. Nate To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message