Weekly Edition
Return to the Press page
| |||||||||||||
|
| |||||||||||||
Is VMware too close to Linux? (The Register)
The Register looks
at VMWare's IPO, which could be in trouble before it happens.
"Writing on his blog VentureCake, Linux specialist Mike MacCana
thinks he might have found a bit of rust on VMware's shining armor. He
claims that VMware's ESX server is derived from Linux, and therefore is not
legally re-distributable as proprietary software."
(Log in to post comments)
Is VMware too close to Linux? (The Register) Posted Aug 17, 2007 23:40 UTC (Fri) by ajross (subscriber, #4563) [Link] The linked article is a big mess of confusion. It's getting its scoop from a blog post here: http://www.venturecake.com/the-vmware-house-of-cards/
That post isn't a whole lot better. But the upshot seems to be that the ESX product boots a linux kernel first to do hardware initialization, and then does a soft reboot into the VMWare kernel, replacing the linux instance.
The sticky part, apparently, is whether such a module constitutes a derived work of the kernel or not. It's really no different than the controversy over proprietary drivers, AFAICT.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 2:52 UTC (Sat) by jamesh (subscriber, #1159) [Link] From the venturecake article (and the referenced emails), it does not sound like the Linux kernel used to boot the system gets replaced as you'd see in a plain bootloader <-> kernel relationship.
Instead the loading process for the VMware kernel modifies the original Linux kernel to talk to the hypervisor from that point on and becomes the service console.
The argument goes that if ESX was using Linux purely as a bootloader, (a) it should be possible to boot without Linux, and (b) it shouldn't need custom kernel modules to load (e.g. use kexec instead).
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 3:11 UTC (Sat) by drag (subscriber, #31333) [Link] It probably uses custom Linux modules because at the time it was created Kexec didn't exist. And again maybe Kexec is so Linux-centric that it's useless for loading non-linux kernels.
Just because something depends on something else doesn't make it derived. It's pretty flaky reasoning.
Is VMware too close to Linux? (The Register) Posted Aug 20, 2007 12:18 UTC (Mon) by paulj (subscriber, #341) [Link] Just because something depends on something else doesn't make it derived. It's pretty flaky reasoning. Actually, that's the most meaningful test a lay software person can easily apply to determine whether a one software work derives from another (IMU).
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 8:12 UTC (Sat) by jsarets (guest, #39560) [Link] My understanding is that vmkernel is not a Linux kernel module. It'ssimply a blob of code that gets loaded and booted by the vmkmod module. It's my understanding that vmkernel doesn't reference any symbols in the Linux-2.4-derived kernel or in vmkmod.
vmkernel may be mapped into the Linux kernel's address space for a
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 3:18 UTC (Sat) by MarkWilliamson (guest, #30166) [Link] Some kernel developers seem to have a slightly skewed idea of what theterm "derived work" means... I think Linus has contributed to it partly by suggesting that stuff which is written specifically to run in the Linux kernel is a "derived work", but if it's ported from another platform it's not.
As far as I can understand, if vmkernel doesn't include source code from
If vmkernel were to link against the Linux kernel, or include Linux
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 4:53 UTC (Sat) by jordanb (subscriber, #45668) [Link] If you write a book, and then I write a book that has same characters as from your book, and they have the same names and personalities, and it's set in the same (or similar) time and place as your book, and it's set to work off of the plot of your book, such that things in it are referenced (as if it's a sequal or a prequal) but my book doesn't have any passages copied directly from yours, and all the text is entirely written by me.
You think there are no copyright issues?
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 5:04 UTC (Sat) by walken (subscriber, #7089) [Link] I can't see any relation between your analogy and the situation at hand.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 5:37 UTC (Sat) by jordanb (subscriber, #45668) [Link] The OP writes:
> As far as I can understand, if vmkernel doesn't include source code from
I contend, using that analogy, that actually copying, verbatim, linux source code is not a requirement for a copyright infringement. I used a derivative book to present a case in which no actual original text (or code) is included, yet it is still an infringing work because the specifics of the book are too tightly coupled with the original.
If a system is so tightly coupled with Linux that it can not function without it, then how is it not derivative?
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 6:13 UTC (Sat) by rsidd (subscriber, #2582) [Link] I contend, using that analogy, that actually copying, verbatim, linux source code is not a requirement for a copyright infringement.]By your reasoning, clean room implementation is not possible. Alternatively, perhaps the concept of plot, storyline, actors and so on doesn't apply to software. A device has specs; a driver, whoever writes it, must follow those specs.
Is VMware too close to Linux? (The Register) Posted Aug 19, 2007 3:01 UTC (Sun) by sepreece (subscriber, #19270) [Link] "By your reasoning, clean room implementation is not possible."
Well, the organization of the code would be covered by copyright. That's probably the best analogy to copying elements of a story. You wouldn't have to copy any actual code, if the structuring was a close enough copy.
Copying functionality alone, but independently implementing it, would not be covered by copyright. To the extent that a driver has to do certain things to make the device work, those things can't be covered by copyright (though the exact expression in code could be). Interfaces generally aren't covered by copyright, at least to the extent that they are functional rather than expressive. If you copied the menu structure of an application, though, you might be in trouble. Needless to say, infringement claims in such areas get to be decided by courts, rather than any obvious and direct application of the law.
There are some court cases that make it appear that anything you need to do to interoperate (interface with) existing software or hardware is functional and non-infringing. How far that principle goes is an open question.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 8:36 UTC (Sat) by hingo (guest, #14792) [Link] I contend, using that analogy, that actually copying, verbatim, linux source code is not a requirement for a copyright infringementSo SCO might have a chance against IBM after all! (Sorry couldn't resist :-)
Is VMware too close to Linux? (The Register) Posted Aug 20, 2007 0:07 UTC (Mon) by MarkWilliamson (guest, #30166) [Link] > If a system is so tightly coupled with Linux that it can not function> without it, then how is it not derivative?
s/Linux/x86/
and suddenly Intel has copyright issues with my code, MS Windows, Linux
It's not that your book analogy isn't good, but it doesn't cover all the
So, the question is, is there a nice interface between vmkernel and
Is VMware too close to Linux? (The Register) Posted Aug 21, 2007 0:31 UTC (Tue) by jordanb (subscriber, #45668) [Link]
> > If a system is so tightly coupled with Linux that it can not function
> s/Linux/x86/
> and suddenly Intel has copyright issues with my code, MS Windows,
Really? Linux seems to work just fine on this SPARC system. What about Linux makes it dependent upon x86?
> Or how about closed source userspace apps? If they use a
They would make that call through the LGPL glibc layer, to a documented, public interface of the kernel no? You think that's no difference than this?
The problem here is that this whole discussion reads like a Slashdot article, where everyone posting here thinks that the law is black and white, that it is simple, that it is intuitive, and that they can speak with authority about it with no legal background, with no understanding of the mechanisms and procedures, and having done no research on the question at hand.
I thought LWN posters were, on the whole, more sophisticated than that.
Contrary to the OP who is 'correcting' the understanding that Linus and other major copyright holders have of copyright, I'm betting that as major copyright holders to an important piece of free software they've received some counsel at one time or another about the copyrights that they hold.
Certainly the GPL was written by very capable lawyers who felt a need to 1) provide a LGPL for linking to incompatibly-licensed programs and 2) specifically use the concept of "mere aggregation" to limit the scope of the license where, presumably, Copyright law was not clear (or was clear but reached further than they wanted the license to reach) on where the GPL obligations would end and by what means it would be made to cover 'integrated' works.
Understanding that:
1) The law is very complicated.
2) It is rarely intuitive.
3) It is not axiomatic, it does not work like a computer program (if boolean then statement else if boolean then other statement, etc). It is subject to interpretation and reinterpretation and then reinterpretation again by the appropriate authorities (who are not internet webforum posters).
4) Lawyers get paid what they do for a reason. It is not a simple thing to decide how a court will act before it issues its ruling, and except in the most simple cases answering a question like 'does this infringe' takes a great deal of understanding and research (as an aside, I find it ironic that the more people know about the law the less willing they are to make declarations about how things work).
Understanding all that, I think the most sensible poster in this whole article is the one who refused to say if he thought it infringed. The reality is that, no matter what the people here want to believe, as far as we're concerned it's just as likely that VMWare infringes as that it does not, lacking any professional legal advice on the matter.
Is VMware too close to Linux? (The Register) Posted Aug 23, 2007 12:29 UTC (Thu) by lysse (guest, #3190) [Link] Hmm. That line of reasoning sounds familiar; I'm sure I've heard something very like it before.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 6:25 UTC (Sat) by kay (subscriber, #1362) [Link] echo "If you write a book, and then I write a book that has same characters as from your book, and they have the same names and personalities, and it's set in the same (or similar) time and place as your book, and it's set to work off of the plot of your book, such that things in it are referenced (as if it's a sequal or a prequal) but my book doesn't have any passages copied directly from yours, and all the text is entirely written by me." >analogy.c
cc analogy.c Looks like books won't compile ... Kay
Is VMware too close to Linux? (The Register) Posted Aug 23, 2007 11:06 UTC (Thu) by NRArnot (subscriber, #3033) [Link] SF books set in someone else's imaginary universe are quite common. If done without the original author's permission, most(?) folks regard it as rude. As far as I know it's not a copyright problem, as long as there is no plagiarism. "Imitation is the sincerest form of flattery"!
And I thought that one of the things that SCO's defeat proves is that you can't claim copyright on a set of look-alike interface definitions either, just as long as the code that implements them is original. Linux would surely qualify as a book written without explicit permission of the original authors, set in the Unix(TM) universe without any plagiarism of the original work.
As for VMWare, I think this company may be the exception that proves the rule that closed-source software will be of lower quality. VMware is astoundingly good, and now that they give away VMware server and VMware player, low-end users don't even have to pay for it.
These days, if you know any Windows user who wants to give Linux a try, or any Linux user who wants to play with other distros, there are no partitioning issues left. Just point them at VMWare player. There are numerous Linux VMs available for download, or any somewhat competent Linux user can brew up his own VMs using VMware server.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 13:21 UTC (Sat) by Zack (guest, #37335) [Link] Most kernel-developers have a rather straight forward idea of what the term "derived work" mean, but the real legal definition of it is slightly skewed.
As another posted pointed out, even a work without a single line of directly lifted code can be a derivative (a translation of a book, for example).
>>So the question is whether he has any solid arguments or evidence to back this up.
He's not payed to be a lawyer, so he uses a very simple metric. It's VMware that is playing the "what can we get away with" game, and he wisely puts the "not a derived work" burden on their shoulders.
Conception of "derivative work" Posted Aug 19, 2007 4:15 UTC (Sun) by butlerm (subscriber, #13312) [Link] Linus' idea of what a derivative work is a complete fantasy, legally speaking. Even if functional dependencies between works had any established legal relevance (which they do not), an independent work of authorship that has not been expressively "recast, transformed, or adapted" from another work is not a "derivative work". 17 USC 101 defines "derivative work" as follows:
[QUOTE]
There might be legitimate reasons why VMWare is violating the Linux kernel copyright in a manner contrary to its license, but any degree of cross-functional dependency (let alone the trivial amount in question here) isn't one of them. Functional dependency alone has nothing to do with whether a work has been "recast, transformed, or adapted" from another work. Functional interfaces, per se, are not copyrightable.
Furthermore the purported "derivative work" formed in memory isn't cast into a tangible medium, and so is ineligible for copyright protection. If there is Linux kernel distribution requiring a license, it would appear to hinge only on the joint distribution of the Linux kernel and other software (of any kind) as a "collective work". If they were distributed independently, that would be a tenuous theory as well.
Conception of "derivative work" Posted Aug 19, 2007 6:37 UTC (Sun) by dark (✭ supporter ✭, #8483) [Link] Hmm, but right now they're distributing a work that includes a copy of a Linux kernel. Surely that is derivative. In order for that to be ok, they have to fall under the "mere aggregation" clause of the GPL. These functional dependencies are at issue there, to determine whether the combination is "mere aggregation", or whether the combination forms a single Program.
Conception of "derivative work" Posted Aug 19, 2007 16:33 UTC (Sun) by butlerm (subscriber, #13312) [Link] Being distributed on the same media does not make works into derivatives of each other, nor does it make the combined distribution into a work that is _adapted_, _recast_, or _transformed_ from the contained works. Such a combination does arguably form what is called a "collective work" - and that may very well be a means whereby arbitrary license terms may be enforced.
That doesn't excuse abusing the language and the law in describing individually authored works as derivatives of each other when there is no such relationship. Is JFS a derivative of Unix System V just because it was developed for that operating system first? Or was it a derivative originally, and then magically lost its derivative status the day it became operational on OS/2? Definitely not.
The right way forward, is as you say, to analyze the situtation in terms of the language of the license. There are no derivative works here.
Conception of "derivative work" Posted Aug 20, 2007 19:16 UTC (Mon) by vonbrand (subscriber, #4458) [Link] Linus' idea of what a derivative work is a complete fantasy, legally speaking. Linus has talked extensively with lawyers about this very issue (it is quite important to him, personally; and there are other stakeholders who'd like any such issues to be cleared up), so I'd be surprised if it really was "complete legal fantasy".
Conception of "derivative work" Posted Aug 21, 2007 21:19 UTC (Tue) by butlerm (subscriber, #13312) [Link] I think the experience with SCO has shown that there is no shortage of lawyers who specialize in that very thing.
What I don't understand is... Posted Aug 18, 2007 6:16 UTC (Sat) by rsidd (subscriber, #2582) [Link] why don't companies who need to put a kernel to such trivial uses simply use a BSD kernel instead? FreeBSD supports most of the hardware linux does, and if all you want to do is initialise the system, surely it will do very nicely.
What I don't understand is... Posted Aug 18, 2007 13:34 UTC (Sat) by Zack (guest, #37335) [Link] Maybe the companies who start out doing otherwise equivalent things with BSD never reach the neccesary clout.
Using the linux kernel is not just a technical decision anymore. If you somehow can link your company or product to "linux", it brings along a lot of free exposure.
So instead of saying, "if they had used BSD earlier on, they would not have this problem", it might just as well be, "if they had used BSD earlier on, they would not be in a position to have this problem."
What I don't understand is... Posted Aug 20, 2007 6:54 UTC (Mon) by gouyou (subscriber, #30290) [Link] In the case of a product like VMWare, I'm not sure that most of the hardware will cut it. The point is that you can today, in the server market, reasonably ask for Linux support; most hardware compagnies are supporting in a form or another Linux (might be binary module, but some form of support is existing). If you want the same thing with FreeBSD ...
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 13:22 UTC (Sat) by nlucas (subscriber, #33793) [Link] Zachary Amsden response on the blog:
http://www.venturecake.com/the-vmware-house-of-cards/#com...
It clears a lot of the BS around the issue, even if you may or not agree with him (I prefer to not take a stance on legal matters like this).
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 17:14 UTC (Sat) by drag (subscriber, #31333) [Link] Well all choosing to not make a choice in a issue does is just assure with 100% accuracy that your wrong. :)
I sorta understand the concept of thinking that says "Oh there are two sides to every issue, and the truth is somewere in the middle". But it's kinda silly if you think about it.
Often two sides is very limiting. There are often 3 or 4 sides to any issue. There can be dozens of different perspectives. When you reduce the arguements down to two possible points of view your realy quite limiting yourself. (not saying that _you_ personally are doing this. I am just saying people in general have this habit). For example, global warming vs not global warming. People often think there are only two sides and thus completely miss the point of what people are saying.
Even if you only have two sides of a arguement, which is rarely the case if you have more then 2 people, it's just as possible that one side is very right and the other side is very wrong.
Also both sides can be sorta right, in which case the true is probably in the middle.
Of course it's fairly common that both sides are completely right, but that there was a misunderstanding and they are arguing about two different things. In this case there is NO 'truth in the middle' since there was no middle to begin with and the whole excersize was completely pointless altogether.
And it's fairly common that both sides are completely wrong. That the truth is so far out there that neither side is ever going to win a argument. Just because the other guy is wrong doesn't mean that your right.
:)
In this case I think the VentureCake guy is way off base. Even if the comment you linked to is technically wrong (which I don't think it is) the Vmware folks have long worked with the Kernel folks in establishing a sort of generic kernel system for providing networking/storage I/O abstractions and such things. Trying to make generic interfaces for Xen, Vmware, and the rest of the stuff.
ESX has been around for a long time now and if any the kernel folks had a problem with how it boots up then I am sure that we would of heard about it from now.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 18:06 UTC (Sat) by JoeBuck (subscriber, #2330) [Link] The "editor" keeps getting his facts wrong, claiming for example that Grub has a different license than the Linux kernel (they are both GPLv2) to make an argument why it's not OK to use a mechanism that completely replaces the Linux kernel by a proprietary kernel. You'd better not reboot your dual-boot system into Windows! A chain of GPL software will load proprietary software into memory.
Is VMware too close to Linux? (The Register) Posted Aug 18, 2007 20:19 UTC (Sat) by nlucas (subscriber, #33793) [Link] I don't take a stance because I already know more about IP law in the US than in Europe and my own country. As non-american and not particularly interested in this things other than the usual curiosity, I know I can never have a more than average opinion about the issue, so I prefer to not think much about it (and I also know this is lwn, where many people actually know what they are talking about).
But if you want my "non-opinion": right now the arguments seem to tend in favour of VMware, but maybe the grey issues are more than enough for the other side to sue, and when this things reach the lawyer level all bets are off.
Is VMware too close to Linux? (The Register) Posted Aug 21, 2007 16:21 UTC (Tue) by drag (subscriber, #31333) [Link] > and when this things reach the lawyer level all bets are off.
Oh hell yes. Lawyers can do a very good job of confusing the issue. They've spent their entire lives with that single goal, especially in U.S. criminal law were if there is any doubt at all to the issue then your innocent.
Is VMware too close to Linux? (The Register) Posted Aug 19, 2007 23:05 UTC (Sun) by mmarq (guest, #2332) [Link] "" Well all choosing to not make a choice in a issue does is just assure with 100% accuracy that your wrong. :) ""
Well, that is the fundamental correct scientific assumption.
But the legal system is entirely about choice. No choice always equal to a 0 position, and positive choice equal to 1. And that varies depending on which side, plaintiff or defendant, you are on the issue. A legal system is always dichotomic.
So the fundamental question on this issue is; Is LF going to sue ?
Because a legal system is based on ' juris-prudence ' and not 'juris-science', a choice is on the balance. 'Prudence' can be changed, but there are always pretty complex decisions that tend to make the issue more complicated to both sides, at the same time that it reinforces the need for more counsel on 'prudence'. Justice far from blind, tends to have a pretty strong corporative position.
From the current 'prudence' position, the positions from some of the Linux development community players are already in big trouble to stand, IMHO...
- from the blog> " This has been constantly found to be true in similar cases, and in fact was vehemently argued correct by many in the open source community when faced with the threat that Linux was a derived work of SCO UNIX because the user / kernel interfaces of Linux were based on the traditional UNIX interfaces (and that argument is well settled). Linux itself crosses the same lines as ESXs vmkernel. Linux can emulate the Windows NDIS layer to load network drivers, and has all sort of shims to allow loading and running binary firmware and drivers. Does this make Linux a derived work on Windows? Certainly not. And if your argument is about to be well, ndiswrapper is technically ok, because it is not part of the kernel, it wraps Windows drivers inside a module, which is separate from the kernel think again. The vmkernel can load drivers from Linux and Windows, and it does so by wrapping those drivers inside a module, which is separate from the vmkernel. You cannot simply choose an position and apply it only when it is convenient. In this case, there is no way for Linux to take any other stand. " -
... which doesn't imply that those positions are fundamentally incorrect, only implying that is 'wise' not to sue.
IMHO, it also makes a strong case for 'DEFENSIVE TECHNICALITIES" like the provisions on GLP v3, and others like DKMS. And IMHO its not at all a different issue. Applyable in these case as well in 'pragmatic' vs 'doctrine', the questions aroused are wrong from a 'good sense' point of view.
Can't i make a similar Linux kernel(like Linux is similar to Unix) full of software patents and protected by DRM measures, to the point i enjoy the fruits of labor of Linux developers, and deny them my contributions ? YES i CAN... Couldn't VmWARE do the same ? YES THEY COULD...
I'm not bashing or defending no one, but it seems to me that Linus [Trovalds] and others positions of the Linux development community, are inflated by a 'NATURAL' corporative view on self-interests, but short centered on the lab and unbalanced by personality conflicts on opinions, which in all does not constitute at all an insult, but only a very common 'and somehow healthy' characteristic of human behavior.
Its repugnant to everyone to see his/her work be taked in advantage by others, and denied any to itself. But that could very well be the case in this VmWARE episode, with the aggravating factor that the legal 'prudence' seems to be adverse. How to prevent that from happening should be the only issue. So before jumping on insulting me, i must say that i'm not making a moral case here, i'm only trying to expose facts to the extense of my ability.
"" Often two sides is very limiting. There are often 3 or 4 sides to any issue. There can be dozens of different perspectives. When you reduce the arguements down to two possible points of view your realy quite limiting yourself. (not saying that _you_ personally are doing this. I am just saying people in general have this habit). For example, global warming vs not global warming. People often think there are only two sides and thus completely miss the point of what people are saying. ""
Very illustrative of what i'm trying to point.
There will be more vmwares cases, some much worst, because things are always changing. Its only NATURAL. For example global warming. What the 'frinking' crackpot was the guy that named 'Greenland'(groenland) to a territory that is a complete block of ice and snow from 8 to 10 months a year ?
Its fair to make the assumption that 'that guy' had a good extensive and reasonable point of view, of the all territory in at least a year time frame, and that we was not pointing to same piece of grass in a south shore... or where there serving some kind of hallucinogenic drugs on board vessels that arrived there, to combat cold and hunger ?...
No historical records of that, and highly unlikely, only that Viking communitys found a land that was really very 'Green' to the point of supporting very well herds of cattle, and good for agriculture, while at the same time, there are historic records of citys and places that are now calculated should be under water in such a scenario, were not !!
How many times were pseudo-scientific hypotheses right in some assumptions and very wrong in others !?... Dynamic is king, and byte hard those that try to reduce things to a static view.
So instead of debating if global warming is a fact or not, the debate should be what are the good consequences and the bad, and here the good seems to outbalance the bad by a large extent, considering the historical facts, and the bad it certainly under an agenda, that is not only psychological but is the ground to impose on a 3th world the denial of industrialization possibilitys !...
So instead of denying vmware posibilityes, IMHO, its better to even facilitate that to happen more often, because change is imperious, but under conditions that not allow them to *pollute the world*... yes namely GPLv3 or other measures... or other 'good sense' measures( not in a envagelization work pro-v3 here)...
And 'good sense' here in analogy, is opposite to the 'bad sense' carbon emission restrictions, when there is contrary evidence, because from 1950- 1980 was when the industrialization grew more rapidly and with much more carbon polluting methods than today, and was when the global temperature went down more rapidly in recent times.
So to restrict any 'supposed' argument is good, but to allow some 'good sense' is to be found.
|
|
||||||||||||