Announcement

Collapse
No announcement yet.

FSF Issues Fresh Statement Over ZFS On Linux With GPL Enforcement

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #71
    Originally posted by duby229 View Post

    I agree with you about BTRFS. I wasn't even trying to do anything fancy with it, just a data drive with a mirror, and I had nothing but problems with it. Mind you this was a long time ago. I haven't given it a fair shot since then and I don't know if the status quo is the same or not. But like you my first impression of it was very bad.
    Really depends on the problems you've had. Which kernel and btrfs-tools versions you were using? The stability improved a lot in 2015.

    Comment


    • #72
      Originally posted by Iksf View Post



      Bit of a side point but I think this perhaps exposes a weakness of the open source model. You both seem prejudiced by your first experiences of the software, with the open source btrfs obviously being far from complete when first made available vs ZFS which was complete before being open sourced.

      Just an interesting note.
      Btrfs offers a totally different level of maturity compared to ZFS. If you need a robust, stable system for production use now, use ZFS. If you know you'll only have a special domain where btrfs's current feature set might fit (and you're willing to patch it, if it doesn't or breaks), it might just do the job. Otherwise I don't really see the point in comparing them at all. What makes Btrfs interesting is its flexibility and licensing compared to ZFS. Eventually it could replace ZFS, but it really depends on your needs, if its sufficient right now. It might also work better for machines that don't have huge loads of ECC RAM. It can also use the shared I/O cache/buffer system in Linux so it might have some other advantages wrt memory use.

      Comment


      • #73
        Originally posted by duby229 View Post

        That's like putting a cart before the horse though, the horse is not going to push. Does this module link to the kernel? If it loads it must. The GPL is not about protecting developers rights on their code, it's about protecting users from having to rely on something they can't modify. The point is that no matter how simple or common the links may be, they are still to the kernel itself. Every binary driver that exists today in fact violates the GPL intended purpose. But people tolerate it on desktop linux because of package management. Personally I'm certain that people will tolerate ZoL as well in the same manner.

        If you don't want to use ZFS, don't use ZFS.

        Don't use BS legalese to try and prevent me from having ZFS just because it can't be licensed to GPL. Besides, GPL specifically exempts ABIs (dynamically linked libraries/modules), and there is no differentiation between kernel ABIs and userspace ABIs/APIs in the license.

        You as a user do not have some intrinsic right to modify software that someone else creates. You do have the right to choose whether or not to use someone else's software.

        If you create software and you want GPL so that improvements remain public, you go ahead and use GPL. You want to prevent me from using some other open source stuff because you want to whine that someone else didn't prefer GPL? Well, go cry in a corner in your room by yourself.

        Originally posted by Iksf View Post
        Bit of a side point but I think this perhaps exposes a weakness of the open source model. You both seem prejudiced by your first experiences of the software, with the open source btrfs obviously being far from complete when first made available vs ZFS which was complete before being open sourced.

        Just an interesting note.
        I agree. I think the problem is that many Linux devs like to claim that something is ready for general use far before it is. Even Linus does this - his kernel "RCs" should be called "beta" and the first several release versions should be the RCs. I had trouble with BTRFS with a basic use case (not recovering from a system freeze) on 3.18.7... I don't remember offhand if they brought out some new BTRFS features with 3.18, but this wasn't 3.18.0 that I was using. I'm sure a bunch of people are going to disagree with me on this (possibly in an overly enthusiatic and mean way) but personally I think this is another huge problem with Linux in general that makes it hard to get into the desktop space.

        I wouldn't have tried BTRFS at that time if it hadn't been called ready by many devs and distros (like OpenSuSE)... Devs should not be trying to convince people that immature code is ready for use. I mean, I wasn't doing this in a business environment, so I wasn't screwed, but I would say that devs should try to hold themselves somewhat responsible for making bad recommendations.

        Some might consider it unfair that I'm not only not giving BTRFS another chance, but am bashing it online, but I consider it unfair that my time was wasted with false claims from people who should know better. Even if BTRFS is much more stable now than when I used it, what does it matter when devs were already telling you to use it when it was bad? This is a "boy who cried wolf" scenario that I did not create. I know a bunch of people here are going to tell me that I should be reporting bugs and fixing source, but the fact of the matter is that unless you want Linux to remain a niche for people who have to report bugs in their filesystem of all things, you should back down on that argument.


        Originally posted by url above
        You'll find zfs.ko automatically built and installed on your Ubuntu systems. No more DKMS-built modules!
        See that .ko? That means this is legal, as it is a dynamic library not based on the Linux kernel. Doesn't matter if DKMS is involved or not; it's still a module. (Assuming also spl.ko)

        (Also, thanks to k1l_ for digging up those URLs)
        Last edited by Holograph; 12 April 2016, 06:11 PM.

        Comment


        • #74
          Originally posted by Holograph View Post

          You as a user do not have some intrinsic right to modify software that someone else creates. You do have the right to choose whether or not to use someone else's software.

          .
          You're completely wrong about this. That is the entire point of the GPL. If you want your users to have the ability to choose, then package management is the proper solution.
          Last edited by duby229; 12 April 2016, 06:14 PM.

          Comment


          • #75
            Originally posted by chithanh View Post
            This is not how it works. The license incompatibility is wholly due to GPL clauses. CDDL doesn't forbid combining with code under other licenses. Whether or not the goal was to prevent use of ZFS code under Linux is inconsequential to this question.
            I didn't mean to imply that the current situation wasn't a discussion about a GPL violation, I am just saying that the purpose of the CDDL is to "protect" code from being used in Linux, so Sun, and therefore now Oracle, must have some legal stand point as well.

            Originally posted by chithanh View Post
            But Oracle as the ZFS rights holder has no legal standing against combining the code.
            I highly doubt that.
            The CDDL wouldn't exist if Sun hadn't seen a way to enforce its main purpose.

            Others, e.g. Linux developers, the FSF, SFC, etc, focus on the GPL because that is the license they care about.

            Cheers,
            _

            Comment


            • #76
              Originally posted by anda_skoa View Post
              I didn't mean to imply that the current situation wasn't a discussion about a GPL violation, I am just saying that the purpose of the CDDL is to "protect" code from being used in Linux, so Sun, and therefore now Oracle, must have some legal stand point as well.


              I highly doubt that.
              The CDDL wouldn't exist if Sun hadn't seen a way to enforce its main purpose.

              Others, e.g. Linux developers, the FSF, SFC, etc, focus on the GPL because that is the license they care about.

              Cheers,
              _

              No, he's right. The thing is, that license allowed everything GPL did and more. GPL is known as a viral license because it infects everything it touches (which is both a good thing and a bad thing) and is the reason that there is an issue here. There is no issue with FreeBSD (as far as I know) because the licenses are compatible (as far as I know).


              Originally posted by duby229 View Post

              You're completely wrong about this. That is the entire point of the GPL.
              GPL applies to statically linked libraries, not dynamic. GPL may be viral but it has limits.

              Also, sidenote, a few people have mentioned this already, but FSF has no stake in this and definitely wouldn't be able to point to any sort of damages caused to themselves or anyone else, so even if this is a GPL violation, all they can do is whine. Fighting about this is not to the benefit of Linux users in the first place. ZFS is open source and that's what matters. GPL specifically does not in this case. This is an issue that needs to be figured out before Linux makes it big on the desktop. (Also, the reason I keep mentioning that is because of the privacy concerns in Windows 10; I've long wanted to see Linux get to the point where my friends using it as well, but now I feel it's significantly more important.)


              (P.S. this will likely be my last post in this thread in interest of letting other people get their two cents in)
              Last edited by Holograph; 12 April 2016, 06:26 PM.

              Comment


              • #77
                The entire topic is about redistribution of both combined GPL and CDDL works combined which is fordidden by the very CDDL license to prevent derivate compliance to GPL when used with Linux kernel. Canonical cannot legally bundle then together regardless what ZoL tried to insunuate and FSF just clarified the reason. As long Oracle aren't interested to clarify CDDL license, it would be safe to avoid bundling OpenZFL with Linux kernel.

                Comment


                • #78
                  Originally posted by Holograph View Post
                  No, he's right.
                  I am not disputing that the current discussion is about violation of the Linux kernel's GPL rather than ZFS' CDDL.

                  I am just pointing out that it is highly unlikely that Sun's lawyers would have crafted a license with the purpose of keeping ZFS (and dtrace) out of Linux without that very license giving them no legal standing to enforce that goal.

                  Originally posted by Holograph View Post
                  GPL is known as a viral license because it infects everything it touches
                  It actualyl does not, but that is a common misconception brough to us by Microsoft in the late 90s.

                  Most libraries used by GPL licensed programs are LGPL, BSD or MIT licensed and had not had to change their license to GPL as "viral" or "infective" would suggest.
                  Th GPL license of the programs' code does not have any affect on the licenses of the libraries, their code remains under whatever license(s) their respective authors have chosen.

                  Originally posted by Holograph View Post
                  GPL applies to statically linked libraries, not dynamic. GPL may be viral but it has limits.
                  There is little difference license wise between different forms of linking.
                  Dynamic linking just makes it easier to keep program and library separated when the library is LGPL licensed and the user must therefore be able to replace it with a modified version.
                  But that is also possible with a statically linked application, provided the user also has the program binary as linkable object code, e.g. as a static library as well.

                  Cheers,
                  _

                  Comment


                  • #79
                    Originally posted by Holograph View Post
                    GPL applies to statically linked libraries, not dynamic.
                    If only...

                    That would be the LGPL. The regular GPL merely contains the option of providing a linking exception clause.

                    Comment


                    • #80
                      Ok, let's make one thing clear: The idea that Sun created the CDDL to prevent stuff leaking into Linux by being incompatible with the GPL is pure FUD and BS.

                      The frequent source of this is a Debian conference where a Sun employee (Danese Cooper) expresses that opinion. The video of said conference is found here:



                      To quote:
                      (Event occurs at 27:26) "Mozilla was selected partially because it is GPL incompatible. That was part of the design when they released OpenSolaris. [...] the engineers who wrote Solaris [...] had some biases about how it should be released, and you have to respect that..."

                      Pretty damning, except that Danese was immediately countered by her boss, Simon Phipps, in the same conference!

                      To quote:

                      (Event occurs at 36:00.) I actually disagree with Danese to some degree...

                      And later on, he expresses his disagreement more vehemently:



                      Even Bryan Cantrill, one of the engineers of Sun, expressed his statements that the CDDL was never viewed by the engineers of Sun as a deterrent for their code getting into Linux:


                      So yeah, one quote without context and framing may make it look like Sun created the CDDL to screw with the GPL, but the whole story paints a very different picture.

                      Comment

                      Working...
                      X