Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
Silicon Graphics

XFS to be released under the GPL 244

Posted by CowboyNeal from the yes-even-more-sgi-news dept.
hedonick wrote in to let us know that SGI will be releasing XFS under the terms of the GPL. It looks like they want to use it to replace ext2 under Linux. Since XFS is a journaling filesystem, this is something I'm personally looking forward to.
This discussion has been archived. No new comments can be posted.

XFS to be released under the GPL More

XFS to be released under the GPL

Comments Filter:
  • There's such thing as IP stack fingerprinting. It's what programs like nmap [insecure.org] do to figure out what the target system is running.

    If Microsoft were to copy Linux's TCP/IP stack, their fingerprint would match as well. It's not easy to change your fingerprint without changing the code to generate what it is you're fingerprinting.

  • > example: how do you prove that Win2K doesn't use some modified Linux IP stack?

    Point well taken, but I think you might have picked a better example... I thought the recent benchmarks showed that the core IP stack was one of the few places where NT actually performed better than Linux. Of course, 2.3 is already far along at addressing the problem, right?


    But my grandest creation, as history will tell,

  • Ahem! (Score:2)

    by styopa ( 58097 )
    What is going on with the comments today. Everyone seems to be saying, roughly, 'What if SGI releases this under GPL and someone other than Linux uses it! That would be bad!' Excuse me!

    Ok, correct me if I'm wrong but I thought the whole point of releasing something open source was so that ANYONE could use it, so long as they published the code and allowed people to modify it. The code, although in a sense given to Linux, is not just for Linux.

    Sure there could be a problem of someone like MS taking the code using it and not publishing it. They would be breaking the law if they did, and if it ever got out then they would be screwed. But assuming that they DID integrate it into Windows, publishing or not, the world wins in one sense. This is because they would be using a very good filesystem. Not only that but it would make projects that allow for reading and writing of slices containing MS would be much easier. Perhaps it will help create more of a standard for file naming.

    Seriously, don't b?tch about another OS using GPLed code. It is seriously uncool.
  • in userland (gcc comes to mind) and optionally in the kernel... to quote LINT from FreeBSD 3.2

    # A math emulator is mandatory if you wish to run on hardware which
    # does not have a floating-point processor. Pick either the original,
    # bogus (but freely-distributable) math emulator, or a much more
    # fully-featured but GPL-licensed emulator taken from Linux.

    Now let's turn the flame throwers off and realize that there are more important things to argue about than licenses. After all, I think most of us use BSD and GPL software constantly. I have a FreeBSD machine, a Linux machine, and an NT machine here at work, and I use 'em all. I know I'm not alone.

  • Re:XFS for FreeBSD (Score:1)

    by Anonymous Coward
    You make the assumption (as many linux guys do) that BSD presents themselves as one OS. They are different OSes, with different goals, and different developers (though there is a lot of code sharing). Linux, on the other hand, presents themselves as one OS because they share a kernel, but really there are tons of Linux distros, and they all have different packaging systems, upgrade procedures, and administration tools. The 4.4BSD derived family all have very similar packaging systems, similar upgrade procedures, but they present themselves as different OSes, not one and the same like linux. Try this: Debian GNU/Linux -> Red Hat Linux -> SuSe Linux -> Stampede Linux -> Slackware Linux -> Caldera OpenLinux -> TurboLinux -> Mandrake Linux -> LinuxPPC -> MkLinux And all the rest of the distributions. And not only that, you can say Linux is ported to _____ hardware platform, but which distributions have linux ports to that platform? Not all distributions have DEC Alpha, Sparc, or PowerPC distributions. And what's this I hear about different alpha specific distrobutions which operate differently? But, for example, NetBSD runs on all of those platforms, and if you get complete CD's, you can install for any of those platforms. With the BSD family, there are three projects, each with a slightly different goal, but when you say NetBSD runs on a Sun3, you don't have to go looking for a NetBSD distribution. So 3 different OSes, or the great number of linux kernel distributions. BSD and linux are both fragmented, but in different ways.
  • example: how do you prove that Win2K doesn't use some modified Linux IP stack?

    Easy. Win2K is up and down like a yo-yo because of it's IP stack on the test site. :)

    Then again, it could just be the 'optimisations' the MS developers put into it. ;) Security through unavailability rules.

  • Big hint: people can already sell free software. It's part of the definition. You know, 'free', as in "you can use this to help yourself in any way, so long as you don't use it to chain someone else."

    Making money is certainly helping yourself.

    -Billy
  • I'm glad we have free licenses like *BSD to go along with the pragmatic licenses like GPL.

    -Billy
  • To "beg the question" means to assume the thing that you're trying to prove

    I thought it meant to assume that something you can prove logically entails the thing you're trying to prove; i.e.:

    1. OS A has threading and OS B doesn't...

    3. Therefore OS A is a faster than OS B.

    Which begs the question:
    2. Is an OS with threading faster than one without? (note: this is hypothetical; I don't even know the answer, and it's probably a 'depends' anyways).

    Correct me if I'm wrong or if that's what you were trying to say and I misunderstood you :)

    chris
  • My RS6000 uses regular UW scsi connectors, nothing proprietary there! However, I don't like the box either. Here's why:

    1: Overpriced.
    - Need I say more?

    2: Poor design.
    - There is NO ROOM inside the damn thing. Granted the tower has some room, the desktop version has no room, looks like a rats nest under the hood.
    - Also, the sliding door on the front conveniently covers the air slots for the two UW hard disks (in our system) which causes most of them to cook. (and crash).
    - Another thing, the box was DOA. This customized server (read 100,000.00$+ box) would just display some useless number on the front, and no boot/video. I called IBM (which was a battle to just talk to them) said to remove ALL the connectors from the system and reconnect. If this didn't work they would send a tech with a new MOBO.

    I dismantled it three times, and no luck. The fourth time, I walked by it and in discust, I smacked it's power switch for the last time, it came up to my surprise and has been fine ever since.

    Back on topic though, AIX is SLOW!! One nice thing though, it is consistently slow, 1 or 1000 users, it always works, and JFS is quite kewl! I have no problems pressing the power switch on a fully operational system, knowing that it will come back without an argument is GREAT!

    Lets hope XFS can do the same for linux!
  • It would be great to see the software dual licensed under an unencumbered license so that FreeBSD could include it.
  • I don't have much experience with XFS, but I do use BeOS and it uses BFS which is a 64-bit journalling FS, and based on that experience I would say that a journalling FS would be a great addition to linux. Not only do you get a stupidly huge max file limit (which doesn't mean much now but will be great when you have a couple of TB+ movies lying around on your hard drive) but the FS is extremly robust... pull the plug in the middle of a large file cp/mv and not only does the machine restart normaly (no fsck) it resumes the copy from where it was when the power went out

    Be also has other FS related goodies like searchable file attributes and mime types, but thats a different story.
  • 1. I do. I use FreeBSD and a journaling FS like XFS would really make my day. 2. Why would you want to? 3. Linux users also wouldn't know how to practicaly apply computing resources if you gave them a manual. 4. Who cares? 5. No it cannot, haven't you read the BSD license?

  • Actually (Score:2)

    by Jordy ( 440 )
    Actually, he is right. I posted this in a comment yesterday [slashdot.org] which shows at the top of the SGI Announces New Strategy and Alliance article. [slashdot.org]

    I also submitted the story yesterday but they said that they were already posting a SGI story and so they wouldn't post mine. :)

    --
  • I think they meant that the code would go into Windows, and they could benefit from it without attributing it to SGI. They would not be sticking to the license, but who would know if no-one's allowed to see the Winslop source code
  • To all the dweebs ranting about how this is old news, the original announcement was that SGI was going to release XFS under an open source license, but did not specify GPL. As I recall, someone from SGI said at the time that the company lawyers had to take a good look at the wording of the GPL to make sure it was compliant with the company's objectives. After all this time, it looks like it passed. Considering SGI must have some pretty expensive legal eagles working for them, this is a major validation of the GPL.

    That a large company like SGI has enough confidence in the GPL to commit one of their major intellectual properties to it is a major news story to me (and one superior to *ahem* certain other proprietary solutions). Maybe even more important than the XFS filesystem itself.

    Programming is 1% inspiration, and 99% plagiarism.
  • Ah, you're right, I should have picked something different

    And like someone (a bit up) said, IP stack fingerprinting can be effective too ..

    Some else mentioned that the GPL is designed to allow people to learn from your code, great idea, that's what makes OSS great stuff. But it would be "immoral" to keep your code to yourself and make a ton of money of a propriatary product after you "ripped" all the good stuff out of a GPLed software.
    I guess my concern is more with the morality of propriatary software (and hardware - think routers,webtv etc.) companies.

    but I'm probably over-paranoid
  • did i recept that right? you think the original author -- from whose code someone realizes some basic ideas -- actually _owns_ those ideas?

    how much of your money you make by doing anything do you hand over to your former school teachers?
  • lfs is no longer in the FreeBSD source tree. NetBSD has fixed this though, check it out.
  • I thought this was announced near 2 months ago. Very spiffy news, anyhow.
  • I have read the GPL which is exactly why I'd never inflict it on someone else.

    And that is exactly why anything I write is under the non-discriminatory BSD license.
  • Sorry, I knew that. I just cut and pasted that list of supported file systems.

  • First Hemos, now Neal.

    How many more postings must we put up with telling us "news" which has already been mentioned and covered in previous posts and comments?

    Malda - it's time to exert some editorial quality control over /.

    D.
    ..is for Defunct.

  • As I said to some other idiot, the BSD code cannot be relicensed under the GPL. That is just as illegal as relicensing GPL code under a BSD license.
  • The fatal flaw with your counter example is that if I wanted the features of package Z, I'd have put them into package Y to begin with :)
  • When can I get it?
  • closed development: This is simply not true. Last month I suddenly turned up on the mailing lists posting code and it has all been welcomed by the community.

    contempt: This is rare.

    hypocrisy: There is current work being done replacing those GNU utilities which are present. Anyway, FSF is just as guilty.

    fragmentation: Free, Net, Open, and BSDi are not the same OS, do not characterize them as the same. 386BSD has been dead for nearly a decade.

    aging user base: I started using BSD three years ago when I was 16.

    innovation: CAM, UVM, and this is only in the past couple months.

    single source: Anyone can resell BSD (Cheap Bytes, Walnut Creek)

    third party: The ports tree (a place where you can install any third party program with a FreeBSD port with one statement) contains nearly 3000 applications.

    CSRG: That was 5 years ago, since then, BSD development has done nothing but grow. BSD is not dying.
  • I can't see it replacing ext2 for the general user distros. I can see value if you're running a server site or some large raid DB but for day to day stuff.
  • Dumbass, that would be illegal.
  • That would violate the license of the Regents of the University of California, just as putting a BSD licensed on GCC would violate the license of the FSF.
  • So does anybody have an idea how long it will be
    until XFS is actually supported by the kernel? Hopefully soon.
  • No, what he said is that a journing fs would probably not make 2.4.0. We may see it in a patch release later and hopefully before 2.6.0.
  • ...most likely against IBM, who is poised to port their journalled file system (JFS, which rox my world) to their "project Monterey" version of a semi-free, semi-commercial x86 UNIX.

    Has anyone ever run comparisons between JFS and XFS? I'd be interested in reading the results of a head-to-head comeptition. I've always been a fan of JFS on my AIX boxen. There's nothing like experiencing a hard power failure and having your UNIX box come back up without so much as batting an eyelash.

  • Actually, according to the IRIX manual for grio, it has much to do with the OS; it will have to do resource evaluation in order to know if the hardware can cope with the request, then some resource reservation stuff; the whole thing is designed for any kind of resource reservation, but it only works on xfs/real-time partitions.
    I suppose this is the reason why they put that disclaimer in the XFS/Linux presentation: it might be too much trouble to introduce this in the linux kernel.
  • that Linus recently announced that the journaling fs would have to wait till version 3, this is good news
  • GPL's got a very nice built in safety to keep that from happening. If you are a company that bases your business model on closed-source software, you take a very very big risk by including GPL'd code. If you are caught, then all of your source code will be exposed to the world because it was based on GPL code.

    So if MS decided that XFS was better than NTFS and decided to coopt it and make a new filesystem, they would risk the entire source code of NT in doing so. It's not so much the likelyhood of being caught that keeps them in line, but the severity of the consequences if they are caught.

    ---

  • Well, I saw it on Tuesday, and I assumed everyone else did too, in this article [slashdot.org] right here on slashdot. Check the comments... One of the score=2 (the highest score of any of the comments in that posting) is entitled "XFS is being released GPL"

    Maybe you should read /.
  • The problem: When a Operating System crashes, the File System is in an unknown and potentially inconsistent state (yes, a BAD thing ;)

    The Solution: Basically keeping a log of all file/disk actions, so we can "restore" and put the system in a known consistent state.

    i.e. A user saves a file in an application
    The File System needs to:
    a) access the meta-file area, and create a new entry (i.e create the file)
    b) allocating space for the file and write the data (commonaly called commiting the data.)

    If the Operating System crashed during step b,
    the next time the user tried to read the file back, it would contain garbage since the meta-data said it was a certain size, but not all of the data was written.

    With journalling, i.e. there would be a step c) Log the file write as complete.

    When the system restarts, the File System would start at the last logged action and work its way to the first logged action undoing all the steps, putting the file system in a known and consistent state.

    Usually the log is implemented in a circular buffer, since keeping logs of EVERY event since startup would quickly fill up disk/memory space.

    Not the best explaination, but it should help.

  • So it seems the point of the GPL is to prevent people from using the code. Sounds an awful lot like proprietary software licenses to me....


    sc
  • Does this mean any changes to the partitoning system that linux uses? Do all those ex2fs partitons in my extended just get converted to XFS if this is implemented? Or does a "new" scheme arise...

    [offtopic]
    Personally, I would appreciate this upgrade if a "new" scheme were implemented, especially if it was similar to the FreeBSD or Solaris way of handling multiple filesystems. Does nobody else think that it's silly that Linux filesystems take up more than 1 logical partiton on your drive? (I speak from the IDE/PC realm) It's a pain in the ass, IMHO, for me to encapsulate 4 to 6 Linux partitons inside an extended partiton just so I can fit 3 other "normal" partitons on the drive that don't require a boot manager to boot. But of course, to even boot Linux in that extended partiton, I HAVE to use a boot manager to boot Linux.

    Wouldn't it be nice to have 6 Linux filesystems sit inside *just one* Linux/XFS logical partiton? Then you can install it with 3 other OSes and just switch the active partiton when you want to boot somewhere. No boot manager needed!

    I'm probably missing something in my logic. Feel free to tell me if I am!
    [/offtopic]

  • First, this is great news because in addition to adding enterprise level (pardon my buzzwords) features to Linux, it also shows SGI's commitment to Linux. Hopefully they will continue to release some of their code (and develop new code) under the GPL.

    I wonder though, at the intent to REPLACE the ext2 filesystem for linux. Yes, I agree that having a journaling filesystem is great for large Linux fileservers, webservers and database servers. Slashdot would certainly benefit from such a system ;-) However, I wonder at the performance hit caused by logging. If I remember correctly, most logging systems work by writing a message to a "log" before changing the filesystem and another after the change is complete. This would greatly increase the disk usage overhead on a desktop (single disk) Linux system like mine. Having such a robust filesystem isn't NECESSARY on my desktop. Is there a way to turn logging off (and remove the performance hit) when using this filesytem. Otherwise I will stick to ext2. Any thoughts?

    Spyky

  • Why would MS (or another vendor) "steal" XFS-GPL, when they could just port it to the Windows NT IFS layer and include the source code on the CD?

    MS might be evil enough to be stealing code from open source software, but if they are, it's purely out of pride, because there's legal ways they could use the code.
    --

  • experience with XFS and JFS (Score:3)

    by jetson123 ( 13128 ) on Thursday August 12, 1999 @07:50AM (#1749877)
    I have been using XFS and JFS-based systems for several years and I'm unimpressed.

    I have more experience with JFS because I have used it in some large applications; I have been running XFS on my desktop Irix machine for a couple of years.

    Here are some observations about JFS; most of those also apply to XFS:

    • It can be quite slow in practice; for example, untarring a large amount of scientific data, JFS took three times as much time as ext2, even though it was running on a faster SCSI disk. I haven't done XFS measurements, but XFS also feels sluggish in practice to me.

    • It only protects file system structure, not file content.

    • It only protects against a small set of failures. For example, hardware failures and file-system related bugs still cause data loss.

    • JFS comes with a LVM (volume manager) and XFS integrates XLV. In my experience, those kinds of systems complicate disk management, increase the risk of file system management mistakes, and make it more difficult to predict performance.

    • Journalling does not guarantee fast recovery. There may still be extensive recovery going on at boot time. IBM's JFS, ironically, often runs on systems that have (unrelated) performance bottlenecks in their boot code that makes them some of the slowest booting UNIX machines in existence.

    • The only time I have actually lost a partition over the last decade was on a journalling file system due to, what appears to have been a software bug in the fs code. Journalling file systems are tricky pieces of software to write.

    Of course, with XFS on Linux, we can finally compare these issues side-by-side on identical hardware and kernels. It will be interesting how XFS holds up.

    XFS has some nice features, and I think it will be a great addition to Linux as an optional file system. Its availability will make Linux much more attractive to some corporate buyers.

    But before adopting it widely, I believe the issues that I raise above need to be looked into and evaluated carefully. I suspect many people don't need the features of XFS, overestimate the safety of journalling file systems, and will get bitten by the complexities and overhead. I hope XFS will not preempt the further development of "traditional" Linux file systems.

  • A quick search of Bloomberg revield nothing. Post a link. Otherwise, these are just more anti-BSD lies.
  • And? I don't think 'standing on someones toes' is the issue here. If Steven Tweedie gets his journaled file system together, then fine - if not, he's missed the boat!

    After all he has been at it for 'some time' now.
  • Just because there will be more than one choice of file system for linux, you shouldn't go nuts about what might happen. Choice is nothing to sniff at, as we get exactly (or close to) what we want.

    In fact, if they are all GPL (and I imagine they would be) then what you may see happen is a merging of technologies into a better product. Say one file system has some sort of shortcoming, but another system doesn't, then you might see a crossover that will result in better products all round.

    I can't wait for XFS, or any other "proven" filesystem (which means that its been tested and works as intended) for Linux, as then I won't have to worry so much about power dropouts. Even with UPS's, sometimes even they fail (as happened to me on the weekend).
  • Well we have a 120G RAID5 an it can take 3 hours to fsck

    I can't wait for XFS
  • If you wanted it fixed, why didn't you fix the code yourself? Frankly, LFS was never really a production quality filesystem, and now with SoftUpdates, I don't miss it in the least.

    If people were rude to you, it's probably because you whined and didn't try to contribute.

  • the discussion on LinuxToday has gone on for some time about the 2.4 kernel. I don't know if the dynamic resizing will be there, but this is what was said about LVM:

    Joe Drew - Subject: Re: Raw I/O, LVM (1999-07-30 05:36:48)
    In fact, both of these features will be added for 2.4. LVM support and raw I/O support are both already in; LVM is undergoing quite a
    bit of hacking activity, but should be ready for 2.4.

  • So the GPL doesn't protect algorithms, just code.

    Correct. This is true of all licenses.

    I can write a program that will mangle code a bit, but produce a program that does the same thing. I've just bypassed the GPL.

    Nope, you've just used that program to create a derivative work, which still falls under the original license.

    There's a program that translates code into plain English. You can use this on crypto source, take it out of the US, and translate it back into source. That would bypass the GPL too, since it reduces the code to raw algorithms.

    Legally, that too is a derivative work. While it may then be in a form that is acceptable for export, it is still a derivative work.


    ---

  • At SGI's Open Projects page for XFS [sgi.com] you'll find that at LinuxWorld, SGI released documetation, data sheets and some sample parts [sgi.com] of the XFS Journalling code (Those parts that are owned by SGI, and which there are no patent issues with).

    In other words, SGI is delivering on their Open Source promise with XFS, even if in only a small way (for the moment).

    Quoted from their own page... "This initial publication of excerpts of XFS Source code indicates SGI's commitment to port XFS to Linux as GPL code. This code can be freely used with Linux, without the worry of violating the GPL terms by mixing copylefted GPL and copyrighted non-GPL code in Linux. It also shows some interesting code (originally developed for IRIX) which implements the XFS journaling techniques."

  • The problem that you are talking about is not a Linux problem, it's a problem with the Intel/IDE BIOS "Standard", which specifies that IDE drives may contain no more than 4 "primary" partitions. AFAIK, SCSI is un-encumbered by this restriction, which is why none of the other UNIces suffer from this apparent "bug". (they all run on hardware where SCSI is the standard way of doing Disk Storage)

    I'm not really qualified to comment on the boot manager issue, but it begs the question, why wouldn't you want to use a boot manager, it makes life so much easier than toggling active partitions in fdisk.

    McA
  • The libraries are under LGPL, man.

    Readline is under GPL, not LGPL.

    Distributing XFS as a plug-in file system separately wouldn't work for M$. You can't use it alone, without the main program. Hence, not "considered independent and separate works in themselves".

    I was not talking about the NT kernel, but some BSD kernel. If you took a BSD kernel, and added some compilation option to build it with the GPL XFS code, by using that option you would create a composite work you must distribute under the terms of the GPL. This you can do, since the BSD license allows you to sublicense the source code under the GPL. And because a BSD kernel is something you can use without the XFS code, it is a separate and independent work from the XFS code.

    ---

  • Just to make sure we're all talking about the same thing here, I'm referring to IBM's jfs, which is part of AIX. I've been using AIX for the last 3 years or thereabouts, mostly AIX version 4 onward. I've got to say that some of your criticisms are unjust. Firstly, yes, the jfs only journals the structure of the filesystem, not the data, but what do you want for a 4Mb journal log?

    Secondly, I have never had any corrupt data on a JFS volume. It's quite neat to be able to hit the power at any given time and not frag the machine. The AIX LVM also is an absolute lifesaver. Yes, it's complex to understand initially (the LVM takes about a day for me to teach in IBM's sysadmin course) but once you know what you're doing it makes life sooo much simpler. Run out of space? Just extend the filesystem. Run out of disk? No problem, just add another one to the volume group (while the machine's still on if you've got hot-swap SSA or SCSI).

    I really couldn't live without it.

  • I think part of the confusion is that most people don't realize that "beg" has a meaning other than "ask for". It can also mean to evade or sidestep. That's the usage in "to beg the question" -- "begging the question" means "avoiding the matter at hand", not "asking for a different question".
  • well, uhh, there is no 8 gig limit on ext2 fs partitions first off... also, on my PII 266 here it does take too much time to fsck my 10 gig /home partition, but it does not take at all near half a day... it takes 4 or 5 minutes.
  • The NT fs arch may or may not be better than vfs, but if no one uses it, it does no one any good. Right now, NT has inadequate choices for file systems, and doesn't seem to be getting any new ones.

  • It doesn't beg any questions at all. To "beg the question" means to assume the thing that you're trying to prove. The common usage of it as "leads to the question of" is completely wrong.

    Sorry, it's a pet peeve.
  • It's not clear whether you are worried about running out of disk space or feeling a "performance hit". Linux is pretty fast (one of the many reasons we all use it). If you are doing desktop-oriented things with your desktop system, you might not even notice a performance hit.

    Drives are so cheap now you might not need to worry about logs taking up lots of space, either.

    It would probably be worth investigating journaling file systems even if your primary use is fairly userly. ext2fs is fast, but it is also one of the most fragile file systems I have ever encountered. IMHO, ext2fs is one of the things about Linux that is mostly likely to deter people from using it in high-end servers. I have lost entire ext2fs file systems irrevocably after power outages or developer-kernel crashes. It's good news indeed that more alternatives are on the way.
  • Well I imagine that the big tip off would be that Windows suddenly start working right...
  • Just in case you didn't know.. XFS and JFS are the same filesystem. They both are written by and licensed from VERITAS Software [veritas.com] (VXFS). It is also the same filesystem that goes into many other commercial unices. Just thought I'd let ya know.

  • Reiserfs doesn't have journalling *now*, but Chris Mason is working on it. Remember that reiserfs hasn't been released. Its feature set isn't set in stone.

    By the way, Reiser has bigger ambitions than just producing a faster fileystem. He aims to change the way people use filesystems. See here [idiom.com] for some insights.

    PS. Hmm, interesting. The slashdot code seems to have removed the spaces in the anchor part of the url. Just jump to "Why Aggregate Small Objects at the File System Level?".
    --

  • The GPL is ugly and immoral. End of story. It also does not promote public domain software. It promotes virus infected software.
  • Wouldn't XFS be able to run under *BSD if it were compiled as a module?

  • An OS with threading will be faster on any computer with more than one CPU, with just one CPU it would probably depend on the quality of the programing
    "Subtle mind control? Why do all these HTML buttons say 'Submit' ?"
  • My only gripe about ext2 is its tendency to fragment. It's nearly as bad as NTFS in this regard. ufs and HPFS do a much better job at resisting fragmentation.

    So, does anyone know if XFS is better at resisting fragmentation than ext2?

    --
    Interested in XFMail? New XFMail home page [slappy.org]

  • Just for the hell of it,

    That would be non sequitur or "does not follow." In essence, you're saying that A leads to B when it really doesn't.

    Begging the question is like:

    Windows is better than Linux, therefore Windows is better than Linux.

    Usually not so blatant but you get the idea.

    Jonathan Wang
  • If you had half a brain, and had actualy *read* the BSD license you would see that you can redistribute the code or binarys however you want. I could take the sum total of the BSD code and put it under the GPL, or sell a propritary version.
    "Subtle mind control? Why do all these HTML buttons say 'Submit' ?"

  • Freebsd is dying (Score:1)

    by Anonymous Coward

    Why is freebsd dying? Each week brings more bad news to the dwindling number of its fans. Despite the mournful wails of the remaining BSD zealots, the leading industry analysts have coolly concluded that BSD has become a dead end as its user base continues to shrink, and ISVs abandon support. What brought about BSD's headlong spiral of death? What lessons can be learned? Here are some of the reasons put forth for freebsd's failure:

    • closed development (real contributions limited to old-boy clique).
    • contempt for others in non-freebsd hacker community.
    • unrepentant hypocrisy (hatred of the FSF while slurping up all the FSF GNU tools in sight).
    • rabid jealousy of the success of others (e.g. the success of Linux and BeOS).
    • fratricidal incompatible fragmentation of BSD community (freebsd, openbsd, netbsd, 386bsd, bsdi).
    • aging user base (gray beard syndrome).
    • lack of true innovation (emphasis on trying to match feature check lists).
    • single source vendor trap.
    • devoid of third party and ISV support.
    • death of CSRG and government subsidies.

    Freebsd is a classic study in office politics and backroom backstabbing. The freebsd ``core team'' has been devastated in recent months, punctuated by the resignation of many of its senior members. From lax attention to security as highlighted in Usenet's comp.risks, to continued financial woes of a certain vendor of freebsd cdroms, the ongoing decline of freebsd should serve well to instruct others in how not to run a software project.

  • Comment removed based on user account deletion
  • Does this mean I would be able to just mount the drive from the
    200 onto a Linux (i386) box and pretty much be all set?


    I think this is exactly what it means. We are also looking forward to being able to do this. I'd love to buy a lower cost Linux box from SGI (although I can't say enough good things about the Origin.) Or put one together myself.

  • (a) Who cares?


    I do, and you should. Just because you don't use *BSD, and I don't use *BSD (although I have OpenBSD installed, and I like what I've seen, I use Linux as my primary OS) doesn't mean that we should not be concerned with the welfare of the *BSD movement. You do not appear to be aware of this, but Linux has gotten a LOT of good out of the BSD developers and BSD code over the years. I know that for a long time (and I believe it still does) the Linux kernel booted and displayed a message to the effect of "This program contains code Copyright (c) the Regents of the University of California, Berkeley ..." You might recognize "Berkeley" as the B in BSD and that snippet (which may not be 100% word for word) as part of the BSD license. Many of the utilities you use every day on your Linux system are *BSD utilities, or derived thereof.


    Be careful who you scorn, especially if you haven't done your research. Remember that competition is a good thing, and that the *BSD people are pursuing a goal, while not identical, similar to that of the Linux crowd.


    (b) You can't integrate other Linux kernel enhancments into BSD


    While this is true (although there may be ways to work around it ?), that does not mean that we should not work for a solution equitable to all parties.


    (c) SGI is getting into Linux. It doesn't give a damn about BSD (as most of us don't)


    I can't speak for SGI, and while I cannot technically speak for "most of us" (as I am sure you cannot), I do not believe this is true. There are many people on Slashdot who care as little for Linux as you seem to care for *BSD. There are also many who feel as I do, using Linux day-to-day but following the *BSD activity and applauding those developers for their efforts, or vice versa.


    (d) Freer license would mean it could get into NT as well


    And what is the problem with that? You (as well as many others) seem to miss the point that a stable and functioning NT would be nothing but a boon to the industry. You rant against Microsoft about the instability and uselessness of their operating system, and then seek to deny them the tools to fix it.


    (e) If BSD really wants it, BSD can GPL itself


    While I don't know the particulars of the BSD license enough to comment on the truth of that statement, I can comment on its invalidity for other reasons. *BSD is under the license that it is under because the authors believed that that license was the one it should be under. Just like you belive the GPL to be superior, they believe the BSD license to be superior. I can see the merits of both. Why don't you release your software under the *BSD license instead of the GPL (if you have any, which I doubt)? That is the same as what you would ask them to do...


    I would like to think that you are just misguided and not a troll. If you are misguided, think before you speak. If you are a troll, go home.


    Ethan

  • I am referring the performance hit caused by requiring 3 writes instead of one to modify any part of the filesystem. One of the major problems, is that the log is usually stored in some physical part of the disk that is not necessarily near the data being written, requiring the head to travel across the disk twice for just one write. In large file servers this can be fixed by hardware means, having multiple disks, etc. But in a desktop this is not the case. You may be right about "not noticing" the hit while doing standard desktop operations (browsing, WPing, etc.), I don't really know how much of a problem it causes, or if it will be significant in this case, its just a though... :-) As for the problems with ext2, I haven't really had any problems. Granted I don't do mission critical stuff on my linux machine, yet... Then again I don't do mission critical stuff :-) I don't think it has any more inherent flaws the HPFS or NTFS.

    Spyky
  • uhh as far as I can tell your friend was running
    /usr/freeware/bin/ls --help

    which says at the bottom:

    "Report bugs to fileutils-bugs@gnu.ai.mit.edu"

    instead of
    /sbin/ls --help

    which yields
    [drek@nessie] ~ > ls --help
    Illegal option -- e
    Usage: ls -RadC1xmnloghrtucpFbqisfLAHMDPSj [files]
    Exit 2

    later,
    -derek
  • Ethan,
    I agree with most of what you say. But theres one
    problem in todays OS-industry. From my expirience
    the main point to get people to use linux or bsd
    is the superior stability. If a company (nowadays
    its MS, but it could have been i.e. ibm) has a
    defacto monopoly and a very strong public image,
    the only way to compete is in fact to be cheaper AND better.
    If one uses bsd-license, he gives away
    all his advantages but gains nothing when it comes
    to easily portable features like filesystems etc..
    So, while I wish the BSD-people the very best, I
    believe in fact that in such wierd circumstances
    as todays IT-industrie, GPL is the way to go for
    many things.
  • I was reading the white paper on XFS when I noticed a section that had some information which might be useful for Mac OS X:

    3.2.5 Attribute Management

    XFS supports user defined Arbitrary Attributes which allow information about a file to be stored outside of the file. For example, a graphic image could have an associated description or a text document could have an attribute showing the language in which the document is written. An unlimited number of attributes can be associated with a file or directory. They can have any name, type or size of value. These Attributes represent the first substantial enhancement to the UNIX file interface in 10 years.
    Wouldn't this be a great way to preserve the Mac OS idea of resource forks? Any ideas?
  • If I were a company releasing source to the public, I would probably not release it under the FreeBSD license because I wouldn't want my competitors to be able to profit from selling it. However, if my goal was simply to get the most widespread adoption of my software, I would likely do a dual license release.
  • I would also like some in depth info on ext3, and what would be nice is a comparison of ext2, ext3, and SGI's XFS. Thanks! rbf ALPHA LINUX POWERED!
  • This is wishful thinking. If someone sued MS for using GPL'd code without releasing their source the court would not (could not) force MS to release their source code.

    The remedy would most likely be a injunction from selling software in violation of the license and money award for damages already done. Just because the license says the code must be released doesn't mean the ct will do so.

    For instance, if the GPL said that violators will be shot and killed, do you think the court would enforce it? Probably not.

  • I'll ignore the "truely [sic] free" part and just concentrate on the license issues.

    I'll quote the GPL, section 2:

    If identifiable sections of that [composite] work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when youdistribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it.

    What this means is the following:

    • If someone were to integrate the BSD kernel with GPL XFS code, and distribute this as a single package, that particular BSD kernel package would have to be distributed in compliance both with the terms of the GPL and BSD licenses at once. In practice this is doable (and the FSF has done it, BTW); it's just a GPL + a BSD advertising clause.
    • If someone were to create an add-on package that allowed people to take the BSD kernel package + a XFS package for it, distributed under the GPL, and compile BSD kernels with XFS support, this could not affect the distribution terms of the GPL code. Stuff like this is done all the time -- look at all the non-GPL free programs that allow you to optionally link them to GPL libraries like Readline (Python and the Haskell interpreter Hugs come to mind).

    Thus, GPL XFS code could be integrated into BSD without tainting its license, since the BSD code is a separate work. What you can't do is distribute the combination under terms incompatible with the GPL.

    ---

  • :SGI's //bin/ls does NOT accept a --help flag, nor does it pop up a GPL notice.

    I DID NOT say /bin/ls! Although I was wrong anyways, I said it was the custom XFS ls that ships with XFS, NOT /bin/ls. (And I was wrong, sorry.)

    It is actually the custom ls that ships with the DMF product. If you execute 'dmls --help' you get a help screen with a GNU license statement.

    My apologies for confusing this with the XFS release to GPL. I'm not an SGI or IRIX user and I botched this up. Sorry.

    But if you are running DMF, give it a try. I understand that SGI is fixing this 'mistake' in the next release.
    -ck
  • :SGI's //bin/ls does NOT accept a --help flag, nor does it pop up a GPL notice.

    I DID NOT say /bin/ls! Although I was wrong anyways, I said it was the custom XFS ls that ships with XFS, NOT /bin/ls. (And I was wrong, sorry.)

    It is actually the custom ls that ships with the DMF product. If you execute 'dmls --help' you get a help screen with a GNU license statement.

    My apologies for confusing this with the XFS release to GPL. I'm not an SGI or IRIX user and I botched this up. Sorry.

    But if you are running DMF, give it a try....

    -ck

  • "But always she's the spectre of uncertainty I first endured, then faded, then embraced..."

    Wow, I didn't realize there were any other people on Earth that listen to Toy Matinee. One of my favorite albums of all time.

  • But I don't think that Apple would switch to XFS this late in the development cycle of OSX; it would be a major change in how the OS handles files and addopting XFS as default would delay X's release.

    OTOH, maybe Apple got some inside info from SGI last year when Apple started to think about licensing OpenGL and will adopt XFS as default.
  • Your suspicions are all the more reason why you should join the fsf cadres. Code can only free so much, you need small arms to put the teeth behind the GPL.

  • The more the merrier. If you have several dozen filesystems to choose from, each with its own unique features (some historical, some futuristic), that is a good thing.

    Of course, most people will probably only use one or two general-purpose filesystems, though the others will find their niches.
  • 1) xfs and jfs are different beasts. There are, I believe, 2 jfs' also. A Journeling filesystem, I believe, usually refers to using a log for metadata only. A log-structured filesystem uses logs for both data and metadata.

    2) WRT write times on a journaled files system:
    if you set your ext2 filesystem to running in synchronous mode, then you may be comparing apples to apples. An ext2 fs is, be defult, not synchronizing to dis, but to cache.

    3) Re: fast recovery: Using a log-structured filesystem does guarantee a recovery time of less then the time needed w/o it. Because if data and metadata is written to a log and the system goes out, the OS is guaranteed to never have to evaluate more of the system status then what is contained within the log(s). As I mentioned before, JFS may only journal metadata.

    4) Log-structured filesystems are extremely fast for small-file queues. Also, log-structured filesystems can easily have features built upon them like defragmentation, snapshots, resizing, etc. because there is no longer any dependancy on the physical aspects of the disk.

    5) Using VxFS w/ solaris, I've never had a critical bug, and I've been able to add space to a volume used by oracle while oracle was running to add tablespace to a production database. That shit's cool. And xfs should be able to do that w/o any problems.

    -Peter
  • If you looked around oss.sgi.com, you'd notice that there are lots of OSS projects. For example, there have been about 20 patches to the Linux kernel, mostly dealing with performance enhancements; some of these are already in the 2.2 and 2.3 trees.
  • What this really tells me is that IRIX comes with a buggy getopt().
  • No you couldn't. The copyright would have to be reproduced in the license. The GPL is viral, ugly, and completely incompatible with anything other than public domain software, which could be put under the GPL.
  • I thought someone in RH's employ (at risk of naming names, someone based in Edinburgh) was writing a journalling add-on for ext2/ext3 ?
  • My apologies for sounding snippy...

    However, I still don't think that posting something in a comment yesterday makes this old news today. I don't read every comment on every article on a slashdot... I read a lot of them, but I don't always have time to read them all. I think that the fact that XFS is being released under GPL is enough for it's own article.

    And I also hate it when people who whine about slashdot. If you don't like it, don't come back.
  • FreeBSD does not have a journaling FS. FreeBSD does support FFS, MFS, NFS, FAT, VFAT, NTFS, Coda, and ISO9660. None of these are jounraling.
  • So SGI will add XFS to the Linux Kernel shortly, great, but what if some company steals stuff from the XFS technology by looking at the kernel source and integrates it into their product?
    I know that the GPL is supposed to prevent that, but the thing is, how do you prove it?
    example: how do you prove that Win2K doesn't use some modified Linux IP stack? Nobody's allowed to see the source so nobody will ever find out right?

    guess i'm just paranoid :-)
  • How many developers will SGI bring to the table? One? or thirty?
    That will realy show me, SGI has less resorces then ever before.
    How much are they going to invest in the future? they have allready paid for XFS.
    What will they do? just maintain the code and stand back? Or will the actively develop on all of the Linux Kernel? or just part?
    That will truly show the heart of SGI.
  • If BSD wants XFS, they'll port it. It'll just reside as an option in the non-free section of the installation.
  • This is a LONG article from SunWorld mag, and you have to go down a fair bit to get to the info about file journaling, and such, but it's a good read:
    Getting to know the Solaris filesystem, Part 1 [sunworld.com]
  • Chew on this D.H. Brown.

    'As for "keeping a log", that is a similar over-simplification. The reference is to "journal file system" a.k.a. a logging file system. This will probably get fixed in Linux within six months, but is currently a definite negative (long fsck times for large storage systems) and just one example of the many criteria we examined in our report.'

    I suppose Linux will move up a notch?

    It will in my book.

    Jim

  • Gosh, all I see is whining about how this is "old news" - I don't recall ever seeing exactly what license XFS was going to be distributed under.

    Regardless, let's showing a little fscking gratitude (forgive the pun) to SGI! They're supporting Linux on their workstations, their developers have always contributed to the kernel, and now they're freeing a huge, important chunk of code that is desperately needed and would be a massive development effort to start from scratch! Hoorah for SGI!

    So what if it's a pre-emptive strike against IBM's journaling file system? More free software is good free software! Now we have more choices, and people can pick what's right for their uses, and may the best product "win", because in the end, we're all going to win. Huzzah! :)

Slashdot Top Deals

Anything free is worth what you pay for it.

Close