This is a read-only archive. Find the latest Linux articles, documentation, and answers at the new Linux.com!

Linux.com

Feature: News

Linux-libre project meets rocky reception

By Bruce Byfield on August 01, 2008 (7:00:00 PM)

Share    Print    Comments   

This year has already seen the second release of gNewSense, the completely free distribution endorsed by the Free Software Foundation (FSF), and the announcement that Ubuntu will have a free software option as part of its installation program. Now, if Alexandre Oliva, a Red Hat employee best-known as a board member of the Free Software Foundation Latin America (FSFLA), has his way, building a 100% free distribution will become easier thanks to his linux-libre project. Unfortunately, the path to freedom, he's finding, is often blocked by politics and a preference for convenience over ideals.

Until recently, software freedom was defined largely by license. But, a couple of years ago, GNewSense developers alerted the free software community to the fact that the Linux kernel contained non-free elements, especially hardware drivers that relied on non-free firmware, by releasing a distribution that stripped out such elements. As a result, not only has the FSF adopted GNewSense as the distribution used in its offices, but this year the FSF released a set of guidelines defining exactly what constitutes a free distribution. In recognition of the newness of this approach, the FSF guidelines clearly state that they are a work in progress, and that "a good faith effort" to remove non-free material is as important as complete compliance.

To help spread the adoption of this new definition, Oliva has developed the linux-libre project to offer completely free kernels. Using a series of scripts, Oliva claims that "maintaining linux-libre takes me two to three minutes a day. Most of the time, no clean-up is required, because the non-free bits in Linux tend to remain in place, and I have scripts that will let me know whether there's anything new that requires manual intervention. It took me several workdays to write these scripts,and I still tweak them every now and then, but it's mostly trivial maintenance these days. When a new Linux release comes up, it takes me just a few minutes of attention to prepare a freed tarball, because I've been accumulating the preparation for it throughout its development cycle."

However, as Oliva has discovered, the use of such kernels is not just a technical matter, but also a political one.

Talking to kernel developers

Oliva knew ahead of time that the Linux kernel community was unlikely to be receptive to the idea of offering linux-libre kernels. "It's not a community that values users' freedom as much as it values convenience," he says. Not only had Oliva firsthand experience from trying to argue about the third release of the GNU General Public License with Linus Torvalds, but "I had reports from developers involved with gNewSense and Blag [another free distribution] who'd tried to submit patches to clean up non-free software from Linux and who had been, let's say, laughed and cursed out of the room."

All the same, Oliva raised the issue last month among Fedora kernel developers. Unluckily for Oliva, the Fedora kernel developers were in the process of resolving the problems involved in maintaining both the standard kernel and Xen versions. "The timing was terrible," he laments. "The idea of adding yet another kernel split was perceived as a risk of bringing all that grief back, so the idea of offering additional kernels built out of linux-libre was vetoed."

Instead, kernel developer David Woodhouse suggested a plan by which non-free firmware would be moved into a separate branch of the kernel source tree and excluded from a normal kernel build. The non-free firmware could be distributed in a separate package by distributions that wanted it, and excluded by those wanting a 100% free distribution or spin.

However, Oliva objects to this solution, arguing that it will simply preserve the existing situation and remove any incentive for vendors to develop free firmware. Just as importantly, "it does nothing to address the problem of other obscure chunks of code in the kernel, often derived from documentation under non-disclosure agreements or from non-free drivers that don't offer permission for the code to be distributed under the GNU General Public License." Each distribution would therefore have to keep a separate eye out for such non-free code. By contrast, using the linux-libre kernel could avoid this duplication of efforts, Oliva suggests.

Fedora Freedom

At the same time, Oliva began promoting his ideas in the Fedora project by creating a Fedora Freedom project. True, the Fedora licensing guidelines specifically exclude non-free firmware from consideration. However, given Fedora's original goal "to build a complete, general purpose operating system exclusively from free software," to say nothing of the enthusiasm in the project for Greg DeKoenigsberg's suggestion that "Freedom is a feature" be used as a project slogan, Oliva hoped that the idea would find some welcome.

In practice, though, Oliva was disillusioned almost immediately. A kernel built from linux-libre, he was told, was unacceptable because it could not be trademarked by Fedora, although a free spin might be welcomed.

In recent weeks, Oliva has turned to the Fedora list to recruit support for developing a free version of Fedora. But here too the reception has been chilly. Not only has Oliva had to explain and constantly defend the new definition of software freedom, but part of the discussion has degenerated into a revival of the old dispute over calling the operating system Linux or GNU/Linux, with his detractors making clear that they care little for ideals compared to personal convenience. Nor, despite a few posters who support Oliva's views, has the discussion engaged more than half a dozen people. However, one of those is Oliva's fellow Red Hat employee Alan Cox, who, as much as anything else, is annoyed at Oliva's insistence on arguing every point in the discussion in exhaustive detail.

Asked if Fedora Freedom is still worth doing despite this lukewarm reception, Oliva replies ruefully, "That's a question I ask myself every time I face this kind of hostility." For now, Fedora Freedom remains a proposal that, despite the small amount of work necessary, seems unlikely to be included in the next Fedora release.

Dealing with disappointments

For Oliva, the experience appears to have been disillusioning, not just because of his free software idealism, but because he thought he had a reasonable approach. "I expected negative reactions to the idea of replacing the non-free kernel," Oliva says, "but I thought adding an alternative 100% free kernel built out of 100% free sources was going to be a no-brainer."

On a personal level, the reception at Fedora is especially disappointing. "I work for Red Hat," he says, "and I'd really like to be able to recommend to people out there a distro that the company I work for helps create. It was quite disappointing that a community that set out to create an operating system exclusively from free software would not permit anyone to install or distribute a system without also installing or distributing non-free software."

If Oliva's experience is any indication, the technology for a totally free operating system is in place, but much of what calls itself the free and open source software community has still to catch up with the idea. Instead, many seem content with a partly free, partly proprietary operating system simply because that is what they know or are used to.

Bruce Byfield is a computer journalist who writes regularly for Linux.com.

Share    Print    Comments   

Comments

on Linux-libre project meets rocky reception

Note: Comments are owned by the poster. We are not responsible for their content.

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 76.74.205.122] on August 01, 2008 07:45 PM
Heh. I wouldn't be using this anytime soon - see no point in limiting what I can do :)

#

Re: Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 82.154.65.186] on August 01, 2008 07:59 PM
So just because it would limit you ( or wouldn't use anyway), you wouldn't like having a choice at install time between one kernel and the other? I'd also like a global worldwide ban on tobacco because I wouldn't be the one smoking anyway since it's bad for my health, so I see your point completely.

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 149.254.192.195] on August 01, 2008 08:11 PM
Currently I use Ubuntu. I have an nVidia graphics card, a webcam that uses proprietary frimware, and a wireless card with the same issues.

They bring instability to my system. I do not find it convenient.

Shortly I will be swapping out my nVidia card, buying a wireless network adaptor that doesn't require firmware, and ignoring the webcam.

These devices all seem to have one big problem: no one can fix them but the manufacturer.

Sure hardware devices often have onboard firmware that doesn't need uploading and the reason to move to uploading firmware is to allow post manufacture updates. Nice theory, doesn't work because no one can update the firmware and manufacturers are only interested in updating it if it saves them in returns, not if it functions kind of sort of enough of the like time to not like cause a problem much.

Open firmware would save so much hassle because in the end the user needs power over their computer and the more power they have, the better it's likely to work for them.

#

There'll always be people who appreciate truly free software

Posted by: Anonymous [ip: 82.192.250.149] on August 01, 2008 09:53 PM
Most users are just "consumers". They use GNU/Linux because they don't have to pay for it, nor for OpenOffice, Gimp, mplayer, Firefox, Apache &c. If Microsoft gave away Windows and MS Office they'd switch to that in a heartbeat.

Those of us who understand why it's important for software to be free will always want free software.

Get away from the mentality that equates success with popularity.

#

Re: There'll always be people who appreciate truly free software

Posted by: Anonymous [ip: 205.208.133.107] on August 02, 2008 10:26 PM
Speak for yourself, I use the best tool for the job. If the best tool happens to be free so the much the better, but if the free tool can't compare to the proprietary "evil" solution than I use proprietary "evil" software. It has nothing to do with convenience and everything to do with the free tool not providing the functionality that I need.

The GIMP is a prime example of people expecting others to use software that simply doesn't provide the functionality that is needed to perform a job. The stable branch currently is _useless_ for anything apart from low bit-depth images. When the Gimp dev's finish GEGL and give us the basic functionality that Photoshop has had since version 6 and Silicon Grail graciously added with their branch of the GIMP back in 1998 that wasn't merged into /trunk through a political decision much like those outlined in the article I may be able to use The GIMP, but as it stands for anything in the high bit depth range I have to use Cinepaint or Krita both of which have their drawbacks when compared to the proprietary solution.

I admire the ideals that you hold dear, but if you want me to adopt those ideals, please improve the solutions that you are proposing.

#

Re(1): There'll always be people who appreciate truly free software

Posted by: Anonymous [ip: 147.62.42.119] on August 04, 2008 12:20 PM
[QUOTE]Speak for yourself, I use the best tool for the job. If the best tool happens to be free so the much the better, but if the free tool can't compare to the proprietary "evil" solution than I use proprietary "evil" software. It has nothing to do with convenience and everything to do with the free tool not providing the functionality that I need[/QUOTE]

Eh? This is exactly the consumer mentality described above. It is very much your convenience defined by what you decide is the best tool for you to complete the job.

The whole idea of Free (as in Freedom, not beer) software is based on collaboration. Sure some things will not work as you want them to. In that case, please join the community to build a better solution. If you can't code, you can certainly test and submit bug-reports and feedback.

If you can't help bridge the gap between the GIMP and Photoshop, you have little right to comment on the deficiencies of the GIMP. People like you should most definitely stick to proprietary software - as all you care about is your personal convenience to "get your job done" using what, in your opinion, is the "best tool for the job".

Contrast people like yourself with the people behind Scribus. Those people behind Scribus could very well whinge like you and complain about the lack of DTP software on GNU/Linux. Note however, that they chose to do something about it instead.

#

Re(1): There'll always be people who appreciate truly free software

Posted by: Anonymous [ip: 147.206.4.254] on August 04, 2008 08:17 PM
"Speak for yourself, I use the best tool for the job. If the best tool happens to be free so the much the better, but if the free tool can't compare to the proprietary "evil" solution than I use proprietary "evil" software. It has nothing to do with convenience and everything to do with the free tool not providing the functionality that I need."

Well, it depends how much you "need" it. If you are tied into a specific app because of contractual arrangements, file formats demanded by a client, etc., then maybe there isn't much you can do about it. But in many if not most cases, the free app and the proprietary app do almost all of the same things, each adding a few of its own features. Are the extra features in the proprietary app *really* that necessary? Usually not. What we need is for more users to value freedom, and basically adopt the attitude "if it isn't Free, it doesn't exist" as much as possible. Note that I said "as much as possible" - I don't think there is anything wrong with using proprietary software to do a needed job that truly cannot be done with Free software.

I think that freedom *absolutely* is a higher priority than technical excellence. We can certainly take a Free but less functional app and make it better. We cannot take a non-Free but more functional app and make it Free.

#

Quote about BLAG incorrect

Posted by: Anonymous [ip: 10.22.22.62] on August 01, 2008 11:36 PM
I'm not sure where the error came from, but BLAG was never "laughed and cursed out of the room" as we never submitted anything upstream nor had any discussions about this with upstream. I didn't see much point to it--they certainly knew what was there, more or less, and it would work it's way upstream if they really wanted it.

In fact, I don't even think I mentioned the project outside of the blag namespace until Oliva came along. At that point there was quite an explosion about the issue on fedora-devel, with lots of ugly FUD.

Alan Cox had mentioned something along those lines too[1]:
===========================================================
Alan Cox wrote:
> On Tue, Mar 25, 2008 at 03:07:44AM -0300, Alexandre Oliva wrote:
>> Considering how much effort gNewSense, BLAG, Dynebolic et al put into
>> it, I don't get the impression that upstream is interested in that.
>
> Upstream has encountered nothing but extremists, nobody pragmatically
> interested in working through the needed changes one driver at a time with
> the maintainer to keep them working nicely. This has left a bad taste
> to say the least.
===========================================================

It's interesting those three distros are getting dragged into it as I don't think *any* of them (and BLAG and dynebolic certainly not) had gone upstream.

I just cleaned out the kernel and made announcement to the BLAG lists. Oliva took a script I did (which was based on one I got from gnewsense), greatly improved it, and is currently building new kernels everyday with it (builds for Fedora 7, 8, 9 and rawhide!). He is doing an *excellent* job of this, and I'm very happy about that. :) (not to mention quite relieved- it's a task I didnt really like...)

I try to stay out of the debate as much as I can, but BLAG seems to get dragged into it in threads and articles with inaccuracies, so I feel I need to correct those.

-Jeff Moe, BLAG maintainer

[1] http://article.gmane.org/gmane.linux.redhat.fedora.devel/80333
Alan Cox has also spread other disinfo about BLAG and -libre, I don't know why.

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 127.0.0.1] on August 01, 2008 11:54 PM
Jeff and Bruce, sorry about the misinformation, my sentence didn't come out quite right. I heard, and I thought it was from you (among others), that gNewSense developers had tried submitting patches upstream, and kernel developers reactions confirmed patches to remove non-Free bits had been submitted before. I may have connected the dots that shouldn't be connected, when I assumed gNewSense had been behind that.

As for mentioning BLAG, it was there because IIRC I had got some of the reports in the form of comments from you, not that you had gone upstream yourself. But I realize I wrote something else. Sorry, mea culpa :-(

--
Alex Oliva

#

Re: Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 10.22.22.62] on August 02, 2008 12:03 AM
No problem, and thanks for the kernels!

I should point out to fedora users that they should ignore the "it won't work on my machine" type of FUD as it probably *does* support all your hardware. Only like 30 drivers are removed, many of those obscure. The only common ones would be some 3D stuff (but intel 3d is still fine) and e100 network driver (but you can just use eepro100).

Anyway, if you're running Fedora 7, 8, 9 or rawhide give the kernel a test. Worst case you reboot into your other kernel. No big deal.

-Jeff, BLAG maintainer

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 66.189.107.166] on August 02, 2008 05:14 AM
linux-libre is very important because it highlights to the users what is truly open source and what is binary blobs. It's very important for anyone concern about security since you can know exactly what is in the code.

I hope this will drive people to make more source available under the gpl and encourage the Linux developers to give users an easy way to compile a truly gpl compliant kernel. If you care about choice you should support linux libre as it gives you the choice not to be spoon fed mysterious binary blobs. If linux libre did not come around, we'd probably be asked to use more and more blobs in are supposedly free software.

Support truly free software today so users have the option tomorrow! The free software movement has history on our side as one by one the the software of convenience on linux has given way to FOSS!

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 83.253.137.142] on August 03, 2008 08:31 AM
So, let me see if I understand things correctly. If a hardware manufacturer decides to save a few bucks on his hardware design and lets the device driver loads the firmware (as opposed to have it stored in an expensive flash memory unit, or something similar) then Olivia argues that that piece of firmware must be OSS. Why, and why should not firmware that is stored on the hardware be OSS in that case?

Having OSS firmware would also complicate things hugely. I doubt much of this firmware could be compiled using the gcc compiler or x86 assembler. So, in order to build a kernel with OSS firmware components a range of different compiler and linker tools would have to be applied, and presumably distributed with the kernel is source code form (such that they can be built on the various platforms that linux should be compiled on). And then there is the even bigger question: who will maintain these OSS firmware components? Yes you can change them to your liking, but can you really? Firmware is generally fairly hardware oriented and to do anything useful with the firmware you would probably need a pretty complete description of the hardware. Writing device drivers is considered by many to be complicated enough. Here we are talking about writing the interface for the drivers, which can be even more complicated.

My opinion is that the linux-kernel should remain a kernel and not turn into a firmware project. Kernels speak to hardware and modern hardware incorporates software (firmware) to make interfacing easier and quicker, and I therefore see no reason to require firmware to be OSS simply because it is loaded by the kernel.

To one of the previous posters: Firmware that is loaded by a driver is no more buggy than firmware that is loaded by the hardware from memory at boot time. Note also that most wifi adapters use firmware in one form or another. Best of luck finding a wifi adapter and webcam that does not use firmware (you may find adapters that loads the firmware of on-board memory but they still use firmware).

#

Re: Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 127.0.0.1] on August 04, 2008 08:26 AM
I don't know who this Olivia is, but she seems misguided. Or perhaps whoever's putting words in her mouth is :-)

There's little ethical difference between denying customers control over their hardware. It makes little difference if it's accomplished through Operating System Software or through software that's pre-loaded by the vendor into non-volatile storage on the device itself. If it can be modified, and the vendor benefits from that, why shouldn't the benefit be extended to the customer as well? The only difference I can see is that, by moving the software to the operating system, the vendor gets even more benefit, out of not having to put in the non-volatile memory, and can also attach a EULA to the software that the user must agree to before being entitled to use the device. This EULA may create other ethical, moral and social issues, of course.

The narrow view of firmware as "software that doesn't run on the main processor" won't last very long. More and more workloads are being pushed onto co-processors (crypto, TCP offloading, GPUs, parallel processing, SoC with heterogeneous cores, etc), and if we were to readily sacrifice our freedoms just because some piece of software doesn't run on the "primary" CPU, we'd soon find out a majority of the essential software on a system will be non-Free Software, and we might even not have a "primary" CPU to run our Free Software on.

As for the alleged complications, that's a red herring. There are pieces of Free firmware in the kernel today, and they don't complicate things one iota. The source code is there, and so is the object code compiled with the suitable tools. I don't see why carrying the source code and the permissions along with this object code, so that people can modify it and maintain it should they want to, is any more complicated than whatever it is that the vendor does to create and publish the firmware in object code form only. Of course the vendor is at an advantage if the vendor refrains from divulging specifications to the hardware. We all know how painful that is. Why some people regard lack of such information as a problem for drivers, and not for firmware, is beyond my comprehension.

--
Alex Oliva

#

Re(1): Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 85.91.175.212] on August 04, 2008 10:16 AM
Dear Alex,

I strongly support your action (but I already use debian :)) , although as it seems it won't get far. But what about joining gNewSense ?

Ideals prevail over concrete realisation. And therefore you are prefectly right in your approach (but that you know :)).

I'm very worried when people talk about << it must work >>, << it must be convenient >>, etc. And the same goes everywhere : << I'll protect the environment by buying an eco-friendly car, but I will not sell my car >>.

Fortunately, there are some people like you who still value << ideals >>.

Best wishes.

stF




#

Re(2): Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 127.0.0.1] on August 04, 2008 11:59 AM
> I strongly support your action

Thanks

> (but I already use debian :))

Would you be willing to help linux-libre build freed-ebian packages?

> although as it seems it won't get far.

Well, considering that it's already been adopted by BLAG, gNewSense and Musix GNU+Linux, and soon dyne:bolic and more, in just a few months, I wouldn't say that. I'm quite happy about it, in spite of Fedora's stance. And, hey!, even if Fedora doesn't formally adopt linux-libre, it was its existence that got the firmware separation project moving, and that's something that Fedora definitely wants to adopt, and that might enable 100% Free spins of Fedora down the road. How could I *not* be happy? :-)

> But what about joining gNewSense ?

Erhm... I'm already involved, to some extent, like, offering some guidance in the Kernel Freedom Verification they're carrying out and looking forward to indication of additional material that linux-libre should take out.

I'm also trying to locate people familiar with .deb building infrastructure that could help build binaries for Debian (freed-ebian) and Ubuntu (uhurubuntu; uhuru is an African word that means freedom :-). gNewSense could then use the latter right away, very much like BLAG does with our freed-ora builds, rather than creating their own builds based on older releases of linux-libre.

If you (or anyone else reading this) would like to help, please join us at linux-libre@fsfla.org
http://www.fsfla.org/mailman/listinfo/linux-libre

--
Alex Oliva

#

Nice idea but .....

Posted by: Anonymous [ip: 60.234.206.1] on August 04, 2008 12:34 AM
Well I certainly agree with having the option to install either a totally 100% free kernel OR your standard run of the mill kernel that has some free and some non free components. However, I feel I would probably be someone that purely used the one containing non free components for the simple fact that I need my system to work and I have a hard enough time configuring Linux to run on my box as it is, and thats using the proprietary drivers if they're available.

Were I to buy/build a machine that I knew was entirely supported by OSS then that might be something entirely different, however if I did that I would be limiting myself to hardware that worked with OSS and therefore not exactly being "free".

There is only a certain distance that you can take anything and although I think the option of free OR free/non free is a good idea, I think if it were pushed to the point that only free kernels were available and my system was going to take me even longer to set up I might just drop Linux.

I'm both an I.T. engineer and a developer, plus I have a family, that doesn't leave me much time for setting up my home/home office systems, I use Linux because I love it, enjoy it while I use it and I'm very much for OSS and GPL, however, it should be realised that proprietary and non OSS software has it's place in this world. For example as well as developing in C using the shell I also use Visual Studio to produce Windows programs. It's evil and it costs a fortune but it works well and it makes my life easier when producing software for Windows.

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 211.26.41.88] on August 04, 2008 02:28 PM
I support your kernel project. People, it's not about forcing the removal of non-free firmware from Linux, it's about providing an option for those who want it. I do have some proprietary software on my machine but I recently removed the Nvidia drivers (2d acceleration only now) and I already feel happier. I hope some day to run without any proprietary software on my computer, including firmware and BIOS, and a completely Free kernel is a prerequisite for this.

#

Linux-libre project meets rocky reception

Posted by: Anonymous [ip: 97.119.162.208] on August 07, 2008 05:39 AM
I'm not all that experienced with operating systems, but isn't OpenBSD supposed to be entirely free software?

#

This story has been archived. Comments can no longer be posted.



 
Tableless layout Validate XHTML 1.0 Strict Validate CSS Powered by Xaraya