Linus Torvalds says Android kernel headers claims totally bogus | Hacker News Reader

Linus Torvalds says Android kernel headers claims totally bogus

h-online.com

182 pointsbconway31 comments

Best part: they quoted Linus where he said the whole affair is about some people's need for attention and that he wished "those people would release their own sex tapes or something, rather than drag the Linux kernel into their sordid world."

H-online then made sure the next words they wrote were "Florian Meuller".

What's this a reference to? Or is it just about Florian Mueller being an attention whore?

That's right. The original claim is from Naughton and Nimmer, but it's Mueller who's publicizing and spreading it widely.

Nope. Nimmer had nothing to do with the original claim. He was merely quoted giving some general (and correct) statements on the copyrightable of header files, and said that whether or not Google did anything wrong depends on exactly what they did, which he had not seen so could not say.

For those not familiar with the legal world, Nimmer is approximately to copyright law what Knuth is to data structures.

It's a different Nimmer. (David Nimmer is the real McCoy, Ray Nimmer is just some guy)

http://en.wikipedia.org/wiki/Florian_M%C3%BCller

Seems to be a pretty open-source minded guy. Interesting that he did this move.

Florian Mueller has claimed (repeatedly) that Linux and Android have serious problems with patents and copyrights.

In a recent thread on Linux Weekly News, he was asked to disclose any financial conflicts of interest, and he avoided answering the question:

http://lwn.net/Articles/434587/

Personally, I would not repeat anything he says without careful fact-checking.

I hope this means the Hacker News community will never again promote Florian Mueller's bogus claims and FUD in here.

Discussing something is promoting it? If anything we don't like ever happens, we must not comment for fear of promoting it? When an idea we don't like or agree with is put forward, we should keep silent rather than providing a counter argument?

Although I would see why you might want to limit comments on this particular story, most of the criticism has been ad hominem attacks about which companies the authors have worked for in the past (or work for now).

It hasn't been a victory for rational discourse on either side.

Hacker News is a site where the users control what stories are discussed. By the nature of the site, discussing something is promoting it. I see nothing wrong with trying to limit stories from sources with clear agendas backed by a history of dubious claims. Especially when there are many other stories worth of attention that get ignored.

I hope that I can see a rational discussion about it and this includes both pro and con Android information.

And I'd rather see incorrect information debunked than only seeing pro-Android topics and messages as it usually happens.

It would be nice to definitively shut down this line of attack against Linux. Having references with Stallman and Linus both claiming that using header files doesn't constitute deriving a work seems pretty iron clad to me.

I'm not a lawyer of course.

Linus has also said you can't use the kernel header files to create non-GPL programs: http://lkml.org/lkml/2003/12/5/13

i liked this part more :)

>You are a weasel, and you are trying to make the world look the way you want it to, rather than the way it _is_.

As noted elsewhere, this thread is about creating non-GPL kernel modules and using the internal kernel headers, not the one used by glibc and bionic.

>I'm not a lawyer of course.

Neither is Linus, as people were quick to point out when those old comments of his were first quoted.

The other thing to consider is, who would have standing to sue for their copyrights being violated and would have a desire to sue Google in what would be a risky lawsuit with little upside for the filer even if they win? The sets of "people with standing" and "people with any desire to sue Google for this" seem disjoint.

I don't know anything about the people who initially brought up the potential copyright infringement, but it seems like the only potential beneficiaries of the allegation are folks who would like companies to worry about Android's IP footing.

Well, Android's IP footing is worrysome considering the lawsuits and all. :)

Forgive me if I'm wrong, but wasn't kernel header files almost the exact case that SCO was suing Linux users over?

I don't think it's in anyone's interest to start claiming that header files are copyrightable.

no it was however a fourth for SCO's case though..

Slightly off topic, but I think related: what's the legality of linking to a GPL library from a non-GPL binary, if you aren't distributing both together? (Or even if you do distribute both)

Lawyers representing ffmpeg/x264 seem to actively look to sue anybody that links to x264 from non-GPL binaries (x264 sell a commercial licence based on the fact that you can't link to x264 from non-GPL code). Yet statements like this, and the similar comments from Richard Stallman, seem to indicate that they won't have a case: using their headers does not subject you to the GPL since the headers are not covered by copyright, and if you don't include the library with your program but make it separate then you aren't distributing. At most you are asking your users to create a derivative work at runtime which may be covered by the terms of the GPL, but the GPL explicitly allows users to do what they like, as long as they don't distribute the result.

As far as I know this "no linking" aspect of the GPL has never been tested in court and appears to be very shaky.

This situation is different from ffmpeg. Kernels do not link to your program in the way that libraries do. When using a library, even a dynamically linked one, the memory image of your program is a combined work. As the CPU is evaluating your program, it does not see the library or treat it differently than any other functions in your program.

System calls can be thought of like RPC. The two programs (yours and the kernel) communicate but do not link together in a single memory image.

Yes, as I said the issue are slightly different. My question is: what legal basis does the developer of a GPL library have to sue somebody who released a program linked with that library. If the two only come together to form a combined work on the users machine, at the users command, then how does the GPL apply to me, the distributor of the the non-GPL program? I am not distributing a combined or derivative work. At most I am using the headers and distributing the symbol names, which are not covered by copyright. When I make no copies, how does copyright law affect me?

If the headers were covered by copyright, that would have provided a way for a developer to prevent use of their GPL library by non-GPL programs.

There are different opinions on this, see:

http://en.wikipedia.org/wiki/GNU_General_Public_License#Link...

Your argument applies in the case where you do not distribute the GPL'd library - as zerohp points out, opinions differ on this case, and it hinges on whether your compiled program would be considered a "derivative work".

However, if you distribute the GPL'd library, then it's a lot more clear-cut. "Derivative work" doesn't enter into it anymore - to distribute the GPL'd library itself you must have a license to do so. If the license you're relying on there is the GPL, you must abide by its terms, including those about linking.

It's cool to see Linus being pragmatic here.

actually he is pragmatic pretty much anywhere Linux/programming. Pragmatically speaking, i'd gather all his emails, comments, etc... and made it a point for CS students to study it - not that it is 100% truth or one should agree with everything there, yet it is a coherent pragmatic system and the skills of building such should be studied and developed, like Golden Gate Bridge (vs. say Bay Bridge).

There's a certain concreteness in what you've describe as he's applied it to technology problems. Taking a step (or two) back and examining the system to notice commonalities or abstractions that can enable the current system. He did this with revision control and we have a general tool now embodied in git. He did this with fork and realized that all the attributes of processes could be selectively shared, not just the subset that fork provides, and now there's a much better, general tool, clone(2), that more constructs, forking and threading, can be implemented on top of without them being exceptions and special cases.

I always love it when Linus takes the FOSS extremists down a peg.

We are not dealing with "FOSS extremists" here. This is a focused FUD campaign against Android, coming from people whose only "FOSS extremism" comes in the form of antipathy.

Reply on news.ycombinator.com