From owner-freebsd-hackers Sun Dec 16 17:40:45 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from tele-post-20.mail.demon.net (tele-post-20.mail.demon.net [194.217.242.20]) by hub.freebsd.org (Postfix) with ESMTP id 53C3C37B41C for ; Sun, 16 Dec 2001 17:40:00 -0800 (PST) Received: from ubik.demon.co.uk ([194.222.125.229]) by tele-post-20.mail.demon.net with esmtp (Exim 2.12 #2) id 16Fml3-0003TG-0K; Mon, 17 Dec 2001 01:39:58 +0000 Message-ID: Date: Mon, 17 Dec 2001 01:33:40 +0000 To: freebsd-hackers@freebsd.org, Hiten Pandya From: Tony Naggs Subject: Re: [SUGGESTION] - JFS for FreeBSD References: <20011211120104.28477.qmail@web21101.mail.yahoo.com> In-Reply-To: <20011211120104.28477.qmail@web21101.mail.yahoo.com> MIME-Version: 1.0 X-Mailer: Turnpike Integrated Version 5.01 U Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG In article <20011211120104.28477.qmail@web21101.mail.yahoo.com>, Hiten Pandya writes >hi, > >BTW, i am a first timer at porting a file system... That is okay, few programmers ever port more than one file system. Most useful is to have some experience programming and debugging FreeBSD kernel code. Also useful would be some knowledge of the theory; how journals for a file system work, optimizing disk accesses, etc... also to have experience in debugging or adding features to one. >if the proffesionals think that it is not wise or >useful to port the FS (especially IBM's), it is OK, Does FreeBSD have something like JFS already? No. Would it be nice to have JFS support in FreeBSD? Yes. Has anybody ever said "I would deploy 1000 FreeBSD servers if they ran JFS"? 100? 10? 1? Probably not. As I understand it, the big win with JFS is that after a system restart fsck has little to do, (barring disk media faults). Just syncing up the file system with the journaled logs. So servers with huge file systems boot up quickly. For a JFS implementation to be accepted in this environment the code has to be good, well tested and avoid hassle with GPL. (IMO) >but, just in case, anyone else (more than three >people) I think three people would be a good number. A file system is fairly small, and the parts all rely on each other. Having too many people means lots of effort is spent on sharing out tasks, coordinating, ... rather than coding & testing. The obvious splits are: file-system design & implementation, utilities, testing and documentation. >would like to port this FS to FreeBSD, my target would >be to get it done by September 2002, if we work in a >group... I don't know how hard it is to fit a new FS into FreeBSD, I haven't even done anything with the kernel up till now. So I would not want to timetable anything too strictly. Though having it available for 5.0 (Q4 20002) would be good! >i dont have web-space where i can host this project, >and we would need a mailing list... probably > >freebsd-jfs would help.. >and http://people.freebsd.org/~bsdjfs freebsd-fs is probably good enough for a mailing list, it is fairly quiet most of the time. >but thats only if three of more people are _really_ >interested in porting it... cause as you know... >porting an IBM file system (from looks) is not a >one man job :-) >Sorry, as you know, i am only 15 years old, and my It is good to be interested and enthusiastic. I do not want to put you off contributing, but I wonder what you are looking for. Committing to a project of several months seems quite a big step for someone who also has school and exams to worry about. My impression from reading your posts here, and elsewhere such as freebsd-ia64, is that you are enthusiastic but want some guidance. Can I make some suggestions: Find a small piece of work to do, maybe a few days. This will help you learn the tools and build your confidence, and also get a perspective on the work required for bigger projects. There is only one line about UFS in the FreeBSD Developers' Handbook. Maybe you could write a couple of pages about it for the documentation project. At least knowing how it interfaces to the kernel is an important step in designing and adding a new file system. Meanwhile I can research what is required to implement JFS. Then, say, after Christmas we can get together and try to make a plan for adding JFS to FreeBSD. Regards, Tony To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message