Debian 6.0 Kernel Will Be Free Of Closed Firmware

[makomk]

I think the question in my mind is why doesn't AMD free the current ucode? Is it because it's tied into the digital restrictions management, or is it a trade secret type of thing?

Apparently, releasing the ucode meaningfully as open source would probably require releasing the design of large chunks of the hardware in enough detail that other companies could easily clone it. I'm not sure if even AMD can easily modify the ucode chunks in any useful fashion.

Imagine a time 10 years from now when hardware manufacturers have wisened up, and allocate a gigabyte or more of flash memory on their video cards to support larger, more all-inclusive firmware. Imagine that "free software" device drivers like `radeong' would then devolve into an exercise of calling into the firmware for high-level tasks. If Linus' policy on non-free firmware were to remain unchallenged by then, we would find ourselves with a free "glue" driver in the mainline kernel, along with a non-free firmware blob implementing the vast majority of the functionality.

If this does happen, it'll most likely be as a direct result of Debian's new policy. Debian doesn't care about non-free firmware that's shipped in Flash or ROM on the hardware, not even if it uses firmware signing and DRM to prevent you from reading or modifying it and contains all kinds of horribly user-unfriendly DRM restrictions. They only care about non-free firmware that's loaded onto the hardware by the drivers.

[Remco]

What we consider hardware is mostly software. And if that software is at all possible to be modified, it should be allowed and... well, be. This has always been the idea of FOSS. The lowest level of hardware is actual hardware. This should be open, too. You may have heard of open source hardware.

All progress starts with a compromise. The FSF and Debian have compromised for decades. And they continue to do so. GNU started on a proprietary UNIX. Debian has lots of non-free software in the repositories. This is not strictly necessary, but it is beneficial to the development of the system. You can't fix everything at once. From time to time, they move the goalposts. If they didn't, people would settle with a suboptimal situation, and progress would stagnate.

This is the general idea. I don't have my crystall ball with me, so I don't know whether this particular decision, at this time, will turn out to be a good one. I guess it won't cause Linux usage figures to rise or drop. Maybe it will lead to a replacement for a blob here and there.

[yotambien]

I think the reaction to this (and, by the same token, the announcement itself) is kind of overblown. Debian has been moving in this direction for years, has always been fairly serious about keeping non-free stuff in the non-free section, and they're still going to build alternate install images with non-free firmware and just say that those are "not part of Debian", the same way that the non-free section is "not part of Debian".

Let's be clear. For people already using Debian this decision will have zero practical consequences: it's not like a squad of fanatics are going to strip the needed firmware out of _my_ computer. In that sense you could think the reaction is "overblown". Because yes, none of the users who voiced in this thread their opposition to this move will have any problems surviving the bonfire of the firmware. However, the situation is quite different for new or unexperienced Linux users, as a result of a disastrous strategy conceived under a most puritan and oportunistic idea of what software is, as well as a twisted conception of "freedom".

See, not even the FSF has a clear cut definition of where the dividing line between software and hardware lies:

"The exception applies to auxiliary processors or low-level processors, none of whose software is meant to be installed or changed by the user or by the seller. This can include, for instance, microcode inside a processor, firmware built into an I/O device, or code compiled into an FPGA. The software in such auxiliary and low-level processors does not count as product software".

Why the microcode inside a CPU is any different to that needed for the normal functioning of a GPU remains unknown. On top of that, software "meant to be installed or changed by the user or the seller" is a terribly lousy definition. One sees an attempt to be practical, rather than a strict definition of what is what. With this in mind, if we are going to be practical, let's do so without making user's life any more difficult than needs to be. The problem is that the practicality the FSF has in mind has way more to do with what they realistically expect to be able to control than with actual user freedom. What this means is that so far the FSF hasn't reached the lunacy levels of wanting to remove or replace what is physically impossible to remove or replace. But as soon as the door is perceived to be half open the rules change and what otherwise is considered hardware now becomes software, with so many holes in their logical argument that you could drive a truck through. Again, what about the CPU? It can be updated by the user, so why it doesn't count? Why the firmware living in the CPU is OK while that fetched from Intel or whatever is not? So what if a manufacturer produces a Linux-powered computer physically locked down so no user changes are possible? Would that be OK since we are not "meant" to do anything with it?

The case of radeon's microcode is just scandalous. These people are spitting on the face of everybody who made an effort to bring about the much longed for OSS drivers, from AMD to all the upstream developers and, ultimately, the users. Again, is this microcode "meant" to be changed by the user or the seller, as opposed to the case of CPUs? Has _anybody_ tried to provide a "free" replacement before happily removing it? What, in the name of all that is holy, is gained by removing this piece of code from the main distribution? How is user freedom in any way increased by turning a capable, modern graphics card into a VESA brick? No, seriously, what is the expected outcome of this decision apart from user alienation? Someone explain me how arm amputation is a good idea when (some dude says) you have a scratch in your finger.

Now, all the people thinking this is great because Debian is to remain as "pure" as possible such as to be the "base" for other distributions, that it will make for a fantastic liveCD to test hardware (sigh), that is not meant to be a desktop OS replacement for Ubuntu (!?) and the rest of similarly retarded drivel, ARE ACTUALLY RIGTH. Because unfortunately, with this kind of policy, that's the position Debian will be relegated to.

[Mr James]

Debian is way too obsessive compulsive.

So is Stallman, but without him, were would we be? Windows.....

[squirrl]

Obvious truths

Debian is way too obsessive compulsive.

[Ex-Cyber]

I think the reaction to this (and, by the same token, the announcement itself) is kind of overblown. Debian has been moving in this direction for years, has always been fairly serious about keeping non-free stuff in the non-free section, and they're still going to build alternate install images with non-free firmware and just say that those are "not part of Debian", the same way that the non-free section is "not part of Debian".

[curfew]

Seems like these guys here are those newlywed Windows-converts who still carry the proprietary mentality with themselves. Here's a hint: Linux isn't Windows.

Some distros target the mainstream folks with their mainstream computers that have closed hardware. But that ain't, has never been and most probably will never be the aim for Debian. Yotambien here, for example, is completely lost. What use would it be if all distros chose the same path of ignorance? Someone might ask if there is any point to keep inventing the wheel over and over again.

Debian has always been a distribution with a 'strict' policy to follow the FOSS philosophy. Before criticizing this decision of the Debian devs, one should first explain why he thinks the FOSS philosophy is flawed, and why everyone should rather just go with the flow.

[TwistedLincoln]

This is really great news. Since gNewSense only updates when Ubuntu releases a long term support version, it isn't a practical solution for those wanting an up-to-date distro. Now one can use Debian, and have rolling releases of live media they can use to test to see if hardware requires non-free firmware or not.

To those that think this is a bad idea -- do you support the idea of Free Software at all? Or do you just support "open source" software? If you support Free Software, you'd want to use as little (if any) proprietary software as possible -- and if that means buying a PCMCIA or ExpressCard wifi card to do so, then so be it.

It isn't as if this change makes your computer useless. As others have said, you can always install the binary blobs if you want. Or get new hardware, which is what I would do. Or...(radical idea), use a different distro! I wonder how many people attacking this move by Debian even use it -- if you do, you should certainly be used to having to do a certain amount of legwork on your own. And if you don't -- why do you care? Debian has never been the distro to recommend to new users, so it isn't like this change is going to somehow hurt the overall effort to convert people to GNU/Linux.

[chithanh]

Has anyone written replacement ucode for any chips where the ucode source has been released? If so has it actually proven to be better than the original ucode? I don't see people lining up to write Linux drivers even when HW specs are available, much less ucode for embedded micro-controllers.

There certainly exists dissatisfaction with firmware bugs in the developer community. I cannot say for radeon, but e.g. Intel wireless firmware has a bug which only Intel can fix, and it greatly reduces the usefulness of the hardware to users who experience the bug.

[yotambien]

I think you are talking about the microcode updates. I was referring to the firmware that lives inside your CPU no matter whether you update it or not. I guess the question is: what sort of computers are the idiots who removed support for my GPU using? An abacus?

To add insult to injury, it just happens that this old R300 of mine is one of the best supported cards in Linux by an OSS driver. And it also happens that AMD (and not only AMD but also Red Hat, Novell and a bunch of independent developers) have done and are doing some good effort to improve that support and extend it to newer generations of GPUs, thanks to the release of the relevant documentation--one could say they are making reality the wet dreams of software freedom priests. And now this. The ability to shoot themselves in the foot these people have never ceases to amaze me.

This is a case of shoving "freedom" down users' throats. Only that it solves exactly ZERO problems, and causes as many as the number of users with the "wrong" set of hardware. It will be interesting to know if those responsible for this did not simply use their energies in the most destructive aspects of this ridiculous enterprise. This is, one would hope that, after installation, users will be notified via a nice GUI window that the reason their systems are not working as expected is solely due to the intransigence and zealotry of the developers. At least, in that way they won't automatically arrive at the conclusion that Linux is utterly useless, and maybe, just maybe, will try another distribution.