Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
×

OpenBFS Reaches Beta 37

Bruno G. Albuquerque writes "The OpenBFS Team (part of the OpenBeOS project that aims to recreate BeOS from scratch) has moved the project status to early beta. This means OpenBFS is now a 100% complete (but barely tested) replacement for the original Be File System (BFS). It is a 64 bit, multithreaded and journaled file system that supports unlimited number os extended attributes and has support some database-like functions (like ultra-fast searches based on an index). OpenBFS is written in C++ and is licensed under the MIT license."
This discussion has been archived. No new comments can be posted.

OpenBFS Reaches Beta

Comments Filter:
  • When (if?) this is ported to other OS's (specifically linux) we will have yet another great filesystem. Im starting to see all those 'linux/bsd/whatever' isnt ready for the enterprise arguments slipping away. Still, why a boat needs a OS is beyond me.
    • Well, since it's written in C++, Linus will refuse to accept it in the kernel. I guess since he's not smart enough to figure out how to write kernel-friendly C++ code, no one else is allowed to figure it out either.
      • Re:Great For Choice (Score:2, Interesting)

        by leviramsey ( 248057 )
        Well, since it's written in C++, Linus will refuse to accept it in the kernel. I guess since he's not smart enough to figure out how to write kernel-friendly C++ code, no one else is allowed to figure it out either.

        Just because the current implementation is in C++ doesn't mean that the only implementation is in C++. Assuming that there's some way to define inodes for this fs, writing an FS driver wouldn't be too difficult.

        The tougher issue is the database aspect of this fs, as it can't be hacked into the current VFS system. The best thing I can think of is exporting a few API hooks with some userspace components to allow for this. The userspace part could be done, of course, in C++.

      • Believe-me, it is not just Linus. C++ in the kernel is really a sensitive matter but it *CAN* be used if you know what you're doing. Basically, no exceptions and no pure-virtual methods.

        -Bruno
    • Here [sourceforge.net]. This project is referered to as a "reference source" by the file system team [sourceforge.net] -- whatever that means.

      One thing that bothers me about most filesystems is limited metadata support. BFS is a notable exception [osnews.com]. One hopes that KDE and GNOME developers could use this to improve on the primitive file handling tools desktop users are currently stuck with.

      • Let me clarify that (and I really have to change that text on the page). Once upon a time there was the start of an OpenBFS team. They *WOULD* be using the Linux BFS code as a reference (with the author permission, AFAIK) but they actually never wrote one single line of code. Eventually me and Axel ttook over the OpenBFS team and ended up basing the code on Axel's previous work on the BFS Recovery Tools package.

        In other words, no code has been used from that Linux driver.

        -Bruno
    • I would love to see this in Linux as well. This is why I switched over to XFS. Unfortunately, I quickly found that it wasn't at all like BFS and I am planning on disposing of it very soon. I used to use BeOS, and out of all the things that BeOS had to offer that made the system as a whole so great, the filesystem is what I miss the most. Unfortunately I can't run it anymore because BeOS won't even boot on my system.
    • Re:Great For Choice (Score:1, Informative)

      by Anonymous Coward
      BeFS is more than just a way to store information on a disk, as is the case with extfs, ufs, xfs, etc. BeFS is special because of it allows attributes. Each file is a regular file with 0 or more attributes. Each attribute has a name and a type (integer, icon, string, etc) with data in them. *and*, the attributes can be indexed for database-like searches. (eg find all files with mime_type = "audio/x-ogg-vorbis" and style = punk and year_released
      I should also mention macintosh resource forks, which is basically 2 files stored under 1 name (and slower to access and unindexed), and NTFS streams (file:stream), which are similar, but untyped and unindexed. Maybe LongHorn will change that.
  • I hear this news, and decide maybe it's a good idea to try it out. So I install BeOS 5.04 Dev Edition (from BeOSOnline.de), get it all set up, compile OpenBFS, and transplant it in.

    I restart my system, and lo and behold, it works. I type in 'touch t', and it successfully creates a file called t. However, I then see a box on the screen saying, "BeOS was unable to initialize a swap file." I realize something's wrong - and when I try to create another file, or do anything really, it refuses, on the account that it's a read-only file system. Of course, write support is implemented, but for some reason it was inaccessible to me.

    At the end of all this, I finally manage to corrupt my filesystem, and despite replacing the original BFS driver, it refuses to boot. So I'm going to have to reinstall. :(

    OBOS, sadly, just isn't ready for primetime yet.
    • Note the part where the OBOS team says that the driver is barely tested. B-e-t-a. :)
      • Re:Oh wow. (Score:2, Informative)

        by psicE ( 126646 )
        Actually, we're both wrong.

        I made the stupid mistake of checking out module BeFS. Apparently, I'm not the only one. :D Had I checked out current, it would have worked. And I will now proceed to do that, though it involves reinstalling BeOS. :()
        • Really, you should not be trying this on a production environment. Anyway, there are some issues that we discovered that will be worked on and as soon as I can reliably boot using it on my computer I will make a compiled package and put it up somewhere with instructions on how to install/uninstall it.
      • Pet peeve time.

        Beta software, by definition, should not be barely tested. In fact, beta software should be feature complete, basically working and usable. This means that it has gone through unit level testing, integration testing, and functional tests.

        In other words, the only remaining unknown bugs are those that can only reasonably be discovered by actual volunteer users in real world situations. These users, called beta testers, generally do not have the ability to diagnose or fix any problems they find.

        In still other words, beta software are already candidates for final release.

        This is of course only theory, and organizations do differ on what is considered beta. However, no commercial outfit would release "barely tested" code to beta testers.

    • OBOS, sadly, just isn't ready for primetime yet.

      This is a BETA test of one small part of OBOS. Why on earth would you think OBOS might be ready for primetime after reading this announcement???
  • by psicE ( 126646 ) on Tuesday September 03, 2002 @09:00PM (#4192624) Homepage
    Don't make the same stupid mistake I did!

    If you browse the OBOS source hierarchy, you'll notice a BeFS module. Do not download that one. Strangely enough, the correct module to download is current. Why the other modules remain there, I have no idea.

    If you download the wrong one, you'll get a two-month old driver; it can't write, and it will not take its time in screwing up your BeOS installation.

Work without a vision is slavery, Vision without work is a pipe dream, But vision with work is the hope of the world.

Working...