> Is literally saying that the whole must be licensed under the GPL (note: even whether there are sections that may be considered independent and/or separate and/or derivative and/or whatever), otherwise you just can't distribute it.
You seem to think it's very clear what the "whole work" is and that it must include ZFS.
See my comment of 2 days ago.[0] Section 2 states "...a whole which is a work based on the Program..." which is, as I've already mentioned, defined at Section 0. I'll say again -- at Section 0, it says "a work based on the Program" is either 1) the licensed program (Linux only), or 2) "any derivative work under copyright law", which ZFS is not, so the combination is such a "mere aggregation."
> Just think about it. The LGPL would make no sense if this wasn't the case.
Whether the LGPL makes sense or not is not really a concern of mine, but the LGPL also only applies to "derived works" as well.
> You seem to think it's very clear what the "whole work" is and that it must include ZFS.
And you disagree with that? Not only it is rather evident (Ubuntu distributes both in the same media, and in the same one-click installer), but you even need to admit this point in order to claim "mere aggregation" as you are now doing for the first time.
> See my comment of 2 days ago.[0] Section 2 states "...a whole which is a work based on the Program..." which is, as I've already mentioned, defined at Section 0. I'll say again -- at Section 0, it says "a work based on the Program" is either 1) the licensed program (Linux only), or 2) "any derivative work under copyright law", which ZFS is not
The "whole" here (Ubuntu) is evidently based on Linux since it literally contains it almost verbatim. You are distributing Linux. I still don't get how you can possibly disagree there. You just keep throwing the argument on whether ZFS is derivative or this or that but it is simply irrelevant.
> so the combination is such a "mere aggregation."
And now we finally get to the real core of the argument, where you are finally claiming that Ubuntu distributing ZFS alongside the kernel is just "mere aggregation", the only exception that the GPL acknowledges that would even be relevant here. It is an important exception -- otherwise all of Ubuntu would have to be GPL-compatible -- but notice that it has absolutely nothing to do with whether the parts are derivatives of each other or not, and in fact the very paragraph which introduces this exception actually mentions this _explicitly_.
Which means that, going back to your assertion, it doesn't follow from the arguments you posit that it is "mere aggregation" -- it cannot be decided based on derivative/not derivative.
As I literally said on my very first message on this discussion, Ubuntu claiming this is "mere aggregation" is stretching it. The ZFS module they ship only works with the very same kernel version they ship and viceversa (this for me is already enough to fail the "it is not derivative" claims -- but that's another story). The installer will one-click link the two for you, so that the module loads into _the same address space as the kernel_, and Ubuntu loses functionalities if you don't use the provided ZFS module and use some other filesystem module instead. They could very well just prelink zfs.ko it into the kernel and it would be practically indistinguishable. It is, in my view, impossible to claim this as "mere aggregation".
And notice that Linus claiming "OpenZFS is not derivative" won't make this any different. Even if it was God himself claiming it. The above is just legal tricky, no matter what the legal status of ZFS itself is.
> the LGPL also only applies to "derived works" as well.
And? That is the intention (and the limit) of copyright. To protect distribution of these.
A license for X software could claim that you can only distribute X alongside other software which has an even number of lines of source code and you'd have to comply with it. Not because "other software" is derivative from X or not, but because you have to comply with it to copy X.
> You just keep throwing the argument on whether ZFS is derivative or this or that but it is simply irrelevant.
And you keep saying things like this without explaining them. Perhaps I've failed in being as clear as I possibly can be, and for that I'm sorry, but this is last comment I'll add to this very long thread, because where as I've have provided references to the text, and tried to explain my thinking, you have not.
> The installer will one-click link the two for you
AFAIK this is false, as I believe I've said before. The zfs.ko is prelinked.
> The ZFS module they ship only works with the very same kernel version they ship and viceversa (this for me is already enough to fail the "it is not derivative" claims -- but that's another story).
So, your idea is because a binary module is dependent in some way upon an underlying substrate like a library or a kernel (as opposed to the source being dependent), it is a derivative work? I think that's a fine test to propose as being the test for whether a work is derivative, but you will need to point at some textual basis for that belief. So far, I don't see one from you, and AFAIK the GPLv2 does not propose a specific boundary itself other than that of "derived work."
> It is, in my view, impossible to claim this as "mere aggregation".
Again -- you have to explain "Why?" of this bare assertion. You use language like "clearly" and "evident" and "impossible" when you haven't explained yourself with reference to the GPL text and/or copyright law.
> A license for X software could claim that you can only distribute X alongside other software which has an even number of lines of source code and you'd have to comply with it. Not because "other software" is derivative from X or not, but because you have to comply with it to copy X.
There are boundaries to what are permissible covenants in a copyright license. Copyright is a actually a rather weak IP right, and one example of a very broad exception/exemption would be copyright "fair use". It's fair use to copy API definitions, for instance.
You seem to think it's very clear what the "whole work" is and that it must include ZFS.
See my comment of 2 days ago.[0] Section 2 states "...a whole which is a work based on the Program..." which is, as I've already mentioned, defined at Section 0. I'll say again -- at Section 0, it says "a work based on the Program" is either 1) the licensed program (Linux only), or 2) "any derivative work under copyright law", which ZFS is not, so the combination is such a "mere aggregation."
> Just think about it. The LGPL would make no sense if this wasn't the case.
Whether the LGPL makes sense or not is not really a concern of mine, but the LGPL also only applies to "derived works" as well.
[0]: https://news.ycombinator.com/item?id=35914449