The FSF Wants Microsoft To Do More To Help Fight Software Patents

[audir8]

Sounds almost like abolishing proprietary software is not enough. Sounds almost like we should abolish all of... present day capitalism?

Neither, just abolish software patents.

[Weasel]

Neither, just abolish software patents.

To note: as someone who totally understands closed source software and am not in a campaign against it because it fills different needs than open source, I approve of this. A lot.

Because I know there will be someone who will claim only the FSF zealots are against patents.

Trade secrets are fine, but patents are just cancer that literally steal your freedom. Closed source software does not steal your freedom, because nobody forces you to use it. Not so much with patents, because even if you DON'T read up on the patent and have nothing to do with it, they can still be enforced to take away your freedom.

I've outlined this previously, don't want to go into it again.

[jacob]

Sounds almost like abolishing proprietary software is not enough. Sounds almost like we should abolish all of... present day capitalism?

That has never been RMS's goal but based on what some proclaim here and elsewhere, you would think so.

[jacob]

None of those companies tried to shakedown Linux users the way MS has been for a decade or more. Linux did hurt IBM and Sun in some important markets, and their transition to embracing Linux and open source was a lot quicker and friendlier. There is nothing wrong with someone like the FSF asking MS to atone for it's software patent sins by asking it to help reform/get rid of them. It might be a bridge too far, but the fact that something like the LOT network exists basically means reform is needed. If there is ever another tech crash and patents get sold off again, we'll find out just how badly.

Again, check your history. Sun has NEVER been friendly towards Linux. As for IBM, their attitude towards Linux was initially the same as Microsoft's, they just saw the opportunity earlier and made the good move for themselves a lot quicker. But at any rate, assuming that Microsoft was so much worse that the others (which is not exactly true, and that's not saying anything positive about MS), then the fact that they are changing would be a much greater cause of celebration, wouldn't it?

I think MS has changed under Nadella, and they will still try to protect their market share, but it will be a fair fight now. Software patents are bad, and all of these actions are basically proving so.

The funny thing is that I'm a FOSS radical in many ways, but I've never been against software patents in principle. Or, more precisely, I'm against what is nowadays called software patents, that is broad and vague patents on user features. But I still don't see any fundamental legitimate reason why a nontrivial algorithm could not be patented.

[audir8]

Again, check your history. Sun has NEVER been friendly towards Linux. As for IBM, their attitude towards Linux was initially the same as Microsoft's, they just saw the opportunity earlier and made the good move for themselves a lot quicker. But at any rate, assuming that Microsoft was so much worse that the others (which is not exactly true, and that's not saying anything positive about MS), then the fact that they are changing would be a much greater cause of celebration, wouldn't it?



The funny thing is that I'm a FOSS radical in many ways, but I've never been against software patents in principle. Or, more precisely, I'm against what is nowadays called software patents, that is broad and vague patents on user features. But I still don't see any fundamental legitimate reason why a nontrivial algorithm could not be patented.

I think you should check your history, Sun moved to a subscription model (becoming essentially agnostic to Enterprise Linux), and they wanted Java/ZFS/MySQL/StarOffice etc.. on Linux. Sun and IBM didn't use patents to shakedown users. I think it's fair to say the patent agreements MS made have left a bitter taste, and it might take some time to get past. https://arstechnica.com/tech-policy/...iller-patents/ I don't deny MS the progress it has made, nor it's history. MS has android apps, virtually all have better alternatives. It's a little hard to get excited about open source VS code or .net, when they too have better alternatives.

Patents as an incentive have failed in software, because software is math. Standards are patent free, gzip was designed to be patent free after the LZW patents. Google writes papers about megastore, omega, tensorflow etc.. and has copyrights but no patents. Patents work best when the development costs are greatest, giving inventors 20 years to recoup their investments. They are virtually useless in software, and the LOT network demonstrates this rather well (everyone gets a license if I ever get rid of a patent).

I would make a minimum $100,000 development cost (not including time) for any patent a requirement for filing, and the number of years a patent is issued can be based on the development cost and invention usage. You're either going to innovate and build a business without rent-seeking, or your work should be a part of culture, similar to the copy-left movement.

[jacob]

I think you should check your history, Sun moved to a subscription model (becoming essentially agnostic to Enterprise Linux), and they wanted Java/ZFS/MySQL/StarOffice etc.. on Linux. Sun and IBM didn't use patents to shakedown users. I think it's fair to say the patent agreements MS made have left a bitter taste, and it might take some time to get past. https://arstechnica.com/tech-policy/...iller-patents/ I don't deny MS the progress it has made, nor it's history. MS has android apps, virtually all have better alternatives. It's a little hard to get excited about open source VS code or .net, when they too have better alternatives.

StarOffice and MySQL were on Linux long before Sun acquired them. In fact both started as third party projects (proprietary in the case of SO, open source for MySQL) aimed primarily at Linux. Sun's only merit in that regard was that they didn't expressly drop Linux as a target for that software when they took control of it.

They never really wanted Java on Linux. For a long time Java on Linux was only available as an unofficial port maintained by the Blackdown project, or an alternative reimplementation by IBM. During that time, Sun's products for Java such as the HotJava browser were actually only distributed as self-extracting installers for Windows and Solaris, "just because". It was possible to run them on Linux (not sure why you would want to do that for, but it was possible) but you had to jump through hoops that were unnecessary and served the unique purpose to try to keep Linux locked out. It was only when they decided to make Java open source, many years later, that Linux became an officially supported platform.

As for ZFS, they were adamant that it must never reach Linux. Again, initially it was non-free, but when they finally decided to make Solaris open source, they came up with their CDDL for no reason other than to ensure that it can't be mixed with GPL code. Not that I cared personally, as I never saw ZFS as the crown jewel that its fans claim it to be, but Sun's intent was clear: no ZFS for Linux. Ever. Even if it's itself open source.
[/QUOTE]

Patents as an incentive have failed in software, because software is math. Standards are patent free, gzip was designed to be patent free after the LZW patents. Google writes papers about megastore, omega, tensorflow etc.. and has copyrights but no patents. Patents work best when the development costs are greatest, giving inventors 20 years to recoup their investments. They are virtually useless in software, and the LOT network demonstrates this rather well (everyone gets a license if I ever get rid of a patent).

I would make a minimum $100,000 development cost (not including time) for any patent a requirement for filing, and the number of years a patent is issued can be based on the development cost and invention usage. You're either going to innovate and build a business without rent-seeking, or your work should be a part of culture, similar to the copy-left movement.

By itself, being math doesn't, or shouldn't, make something non-patentable. The general underlying principle is that you can't patent phenomena of nature, but you can patent inventions that apply them to solve a particular problem. For example, you can't patent the electromagnetic force, but you can patent a particular design of an electric motor. By the same token, I think that theorems should be considered as natural occurrences (or discoveries) and thus non-patentable, but algorithms based on them are inventions, and I have yet to see a compelling reason why the general principle should apply in this case.

Obviously I'm talking about a much stricter and narrower definition of software patents than what is usually understood by it. But if it's clear that you shouldn't be able to patent the observed human ear's ELC and masking, I don't see why it shouldn't be possible to patent the MP3 codec which is a specific nontrivial application of these phenomena.

It's also not all that clear that software patents don't generally work. In fact it's precisely because MP3 was patented that we got Vorbis and Opus, arguably superior technologies that serve the same purpose, only better. It's because the MPEG-LA is a patent minefield that we got AV1. It's because GIF was patented that we got PNG. It's also because the RCU algorithm, used by Linux by the way, is patented that DragonflyBSD is trying a different approach with their serialisation tokens that may yet prove interesting and actually worthwile.

So in short no, I'm not convinced that software patents are not legitimate. Saying it would be easier for us if they weren't is not a sufficient reason. I'm also not convinced that they are not useful in driving innovation, including in open source.

[audir8]

StarOffice and MySQL were on Linux long before Sun acquired them. In fact both started as third party projects (proprietary in the case of SO, open source for MySQL) aimed primarily at Linux. Sun's only merit in that regard was that they didn't expressly drop Linux as a target for that software when they took control of it.

They never really wanted Java on Linux. For a long time Java on Linux was only available as an unofficial port maintained by the Blackdown project, or an alternative reimplementation by IBM. During that time, Sun's products for Java such as the HotJava browser were actually only distributed as self-extracting installers for Windows and Solaris, "just because". It was possible to run them on Linux (not sure why you would want to do that for, but it was possible) but you had to jump through hoops that were unnecessary and served the unique purpose to try to keep Linux locked out. It was only when they decided to make Java open source, many years later, that Linux became an officially supported platform.

As for ZFS, they were adamant that it must never reach Linux. Again, initially it was non-free, but when they finally decided to make Solaris open source, they came up with their CDDL for no reason other than to ensure that it can't be mixed with GPL code. Not that I cared personally, as I never saw ZFS as the crown jewel that its fans claim it to be, but Sun's intent was clear: no ZFS for Linux. Ever. Even if it's itself open source.

Sun wanted an Nvidia or AFS-like exception for ZFS on Linux, but it never happened. Java has been a first class citizen on Linux since 1.2, which is about the time the first 2.6 kernel was released (~20 years). To look at how long Java has been on Linux and say Sun never wanted Java on Linux is crazy talk. They also never sued Google for Android, like Oracle did. They weren't perfect, but they weren't anti-Linux like MS was.

By itself, being math doesn't, or shouldn't, make something non-patentable. The general underlying principle is that you can't patent phenomena of nature, but you can patent inventions that apply them to solve a particular problem. For example, you can't patent the electromagnetic force, but you can patent a particular design of an electric motor. By the same token, I think that theorems should be considered as natural occurrences (or discoveries) and thus non-patentable, but algorithms based on them are inventions, and I have yet to see a compelling reason why the general principle should apply in this case.

Obviously I'm talking about a much stricter and narrower definition of software patents than what is usually understood by it. But if it's clear that you shouldn't be able to patent the observed human ear's ELC and masking, I don't see why it shouldn't be possible to patent the MP3 codec which is a specific nontrivial application of these phenomena.

It's also not all that clear that software patents don't generally work. In fact it's precisely because MP3 was patented that we got Vorbis and Opus, arguably superior technologies that serve the same purpose, only better. It's because the MPEG-LA is a patent minefield that we got AV1. It's because GIF was patented that we got PNG. It's also because the RCU algorithm, used by Linux by the way, is patented that DragonflyBSD is trying a different approach with their serialisation tokens that may yet prove interesting and actually worthwile.

So in short no, I'm not convinced that software patents are not legitimate. Saying it would be easier for us if they weren't is not a sufficient reason. I'm also not convinced that they are not useful in driving innovation, including in open source.

Saying patent free software exists because of patents, is a reason for patents being unhelpful not helpful. It's not like newer codecs will become patented again. You're also not looking at Patents and Copyright as a way of offering incentives and furthering innovation. The innovation often isn't in the algorithm, but in it's uses, this is where every example you listed fails in furthering innovation and offering the right incentives. I'm making an economic case against software patents, though I think the case for not patenting algorithms because they're math is pretty good too: abstract ideas shouldn't be patented, they can be copyright when you write them down.

The Apache license and donated patents let patents co-exist with open source software, but as Facebook's BSD+Patents license and MS's past actions shows, patents are a scourge on open source. Even holding defensive patent hordes stops innovations, since people will refuse to build on top of something that is patented.

[jacob]

Sun wanted an Nvidia or AFS-like exception for ZFS on Linux, but it never happened. Java has been a first class citizen on Linux since 1.2, which is about the time the first 2.6 kernel was released (~20 years). To look at how long Java has been on Linux and say Sun never wanted Java on Linux is crazy talk. They also never sued Google for Android, like Oracle did. They weren't perfect, but they weren't anti-Linux like MS was.

The "first class citizen" release of Java 1.2 was a rebranded Blackdown port pushed by Inprise & Sun, with no credits to the Blackdown project and little to zero support. They got bashed in the community for their attitude and although they later apologized, the damage was done. Java on Linux remained mainly a Blackdown project until 2007 when the upstream JVM finally went open source. That was good, but it only came 9 years later.

Neither Nvidia nor AFS have any "exceptions". Nvidia is a completely out-of-tree, nonfree third party driver. OpenAFS, while open source, is not merged in the kernel either, rather there is kAFS, developed by RedHat under the GPL specifically because the IPL and GPL are mutually incompatible. If Sun ever made any efforts to get ZFS into Linux despite coming up with the botched CDDL on purpose to be GPL incompatible, please share your sources or references. I have never heard of anything of that kind and I suspect I'm not the only one.

Yes Sun was less evil than Oracle as far as Java goes, but that's not saying much. Towards Linux they were in fact exactly at the very same level of evil as MS, because, as we shouldn't forget, together with MS they bought the infamous SCO "licences".

Saying patent free software exists because of patents, is a reason for patents being unhelpful not helpful. It's not like newer codecs will become patented again. You're also not looking at Patents and Copyright as a way of offering incentives and furthering innovation. The innovation often isn't in the algorithm, but in it's uses, this is where every example you listed fails in furthering innovation and offering the right incentives. I'm making an economic case against software patents, though I think the case for not patenting algorithms because they're math is pretty good too: abstract ideas shouldn't be patented, they can be copyright when you write them down.

You misunderstood my post. It is a fact, not my opinion, that the Vorbis codec was developed purely because MP3 couldn't be used due to patents. Same thing with AV1 vs AVC/HEVC. In this case, the patent system had served its purpose: it stimulated innovation and the creation of better technologies, and the open source world is none the worse for it.

On the whole subject of patents vs. copyrights, while I'm in favour of patents on specific nontrivial algorithms, I in fact have trouble seeing why exactly should software be subject to copyright. An algorithm is a mechanism, it doesn't matter if a mechanism is abstract or physical. You can't "copyright" a car engine but you can patent it. Logically it should be the same for algorithms. As for the actual implementation, well it's not an artistic creation. Re-implementing the same algorithm doesn't devalue the algorithm itself. So in my ideal world, software (whether source code or binary) would not be copyrightable any more than a mathematical formula is copyrightable. An algorithm should be treated either as a trade secret (in nonfree software) but thus obviously broadly subject to reverse-engineering and with no claims possible against anyone who manages to work it out and reuse it, or it could be patented, with the patent publicly accessible and subject to a reasonable expiry deadline with no possibility to renew it.

[audir8]

The "first class citizen" release of Java 1.2 was a rebranded Blackdown port pushed by Inprise & Sun, with no credits to the Blackdown project and little to zero support. They got bashed in the community for their attitude and although they later apologized, the damage was done. Java on Linux remained mainly a Blackdown project until 2007 when the upstream JVM finally went open source. That was good, but it only came 9 years later.

Neither Nvidia nor AFS have any "exceptions". Nvidia is a completely out-of-tree, nonfree third party driver. OpenAFS, while open source, is not merged in the kernel either, rather there is kAFS, developed by RedHat under the GPL specifically because the IPL and GPL are mutually incompatible. If Sun ever made any efforts to get ZFS into Linux despite coming up with the botched CDDL on purpose to be GPL incompatible, please share your sources or references. I have never heard of anything of that kind and I suspect I'm not the only one.

Yes Sun was less evil than Oracle as far as Java goes, but that's not saying much. Towards Linux they were in fact exactly at the very same level of evil as MS, because, as we shouldn't forget, together with MS they bought the infamous SCO "licences".



You misunderstood my post. It is a fact, not my opinion, that the Vorbis codec was developed purely because MP3 couldn't be used due to patents. Same thing with AV1 vs AVC/HEVC. In this case, the patent system had served its purpose: it stimulated innovation and the creation of better technologies, and the open source world is none the worse for it.

On the whole subject of patents vs. copyrights, while I'm in favour of patents on specific nontrivial algorithms, I in fact have trouble seeing why exactly should software be subject to copyright. An algorithm is a mechanism, it doesn't matter if a mechanism is abstract or physical. You can't "copyright" a car engine but you can patent it. Logically it should be the same for algorithms. As for the actual implementation, well it's not an artistic creation. Re-implementing the same algorithm doesn't devalue the algorithm itself. So in my ideal world, software (whether source code or binary) would not be copyrightable any more than a mathematical formula is copyrightable. An algorithm should be treated either as a trade secret (in nonfree software) but thus obviously broadly subject to reverse-engineering and with no claims possible against anyone who manages to work it out and reuse it, or it could be patented, with the patent publicly accessible and subject to a reasonable expiry deadline with no possibility to renew it.

Nvidia uses special non-GPL symbols that are in-tree. This exception was made just for their driver. IANAL, but almost any piece of code can be incorporated with a GPL kernel in this way. The AFS case is mentioned by Linus here: https://yarchive.net/comp/linux/gpl_modules.html This is before kAFS, but people could link and distribute kernel+openafs binaries together without infringing on GPLv2. You can Google Jonathan Schwartz saying he wanted ZFS on Linux, like on FreeBSD. If CDDL vs GPL could be treated like IPL vs GPL in the OpenAFS case, or like the Nvidia driver, ZFS could have been in the kernel. ZFS isn't CDDL primarily for incompatibility with Linux, it's CDDL because that's what made sense for Solaris and other Sun open source efforts at the time.

Sun was an SCO licensee, but this was in the early 2000's, by the late 2000's Sun wasn't as big as IBM or MS, and desperately trying to figure out how to make well with OSS and Linux to the extent of re-licensing Java from SCSL to GPL, and open sourcing Solaris/ZFS/SPARC/StarOffice/etc..

A car engine isn't usable without being made, and it should be the same with an MP3 codec: you can patent an translation machine, or a audio recorder/player (and there are such patents), you can't patent any abstract algorithms within it. You can copyright a math book full of formulas or a thesis paper, and you can copyright code in the same way. I think your logic is severely flawed, abstract ideas however complex in art or science aren't patented, only copyright. The patent system didn't serve it's purpose for MP3/Vorbis because it stopped innovation until it could be bypassed altogether.

[jacob]

Nvidia uses special non-GPL symbols that are in-tree. This exception was made just for their driver. IANAL, but almost any piece of code can be incorporated with a GPL kernel in this way. The AFS case is mentioned by Linus here: https://yarchive.net/comp/linux/gpl_modules.html This is before kAFS, but people could link and distribute kernel+openafs binaries together without infringing on GPLv2.

The kernel has always had non-GPL symbols and entry points, which are a subset of all the symbols available. Anyone can develop a non-GPL driver using exclusively these symbols. That has always been the case with Linux and it's nothing specific to Nvidia. An exception would be if a non-GPL driver was allowed to use GPL only symbols. Does NVidia do that? I'm pretty sure it doesn't. I remember some company trying to cheat by including a string starting with "GPL" in their module (was it TBS?) and it didn't exactly go well with kernel developers. The same thing applies to AFS.

You can Google Jonathan Schwartz saying he wanted ZFS on Linux, like on FreeBSD. If CDDL vs GPL could be treated like IPL vs GPL in the OpenAFS case, or like the Nvidia driver, ZFS could have been in the kernel. ZFS isn't CDDL primarily for incompatibility with Linux, it's CDDL because that's what made sense for Solaris and other Sun open source efforts at the time.

Sun wanted the kind of exception we talked about above so that they can release ZFS without actually following the rules. They were told that no, it doesn't work that way. They didn't want to open ZFS to the GPL so that it could be used as a legitimate part of Linux, and so it isn't

Sun was an SCO licensee, but this was in the early 2000's, by the late 2000's Sun wasn't as big as IBM or MS, and desperately trying to figure out how to make well with OSS and Linux to the extent of re-licensing Java from SCSL to GPL, and open sourcing Solaris/ZFS/SPARC/StarOffice/etc..

Sun had open source efforts but these were expressly and purposely hostile towards Linux. Their (pipe) dream was to create a parallel community centred around Solaris that would, they hoped, push Linux out of the market. There is nothing inherently illegitimate about that, mind you, no-one says that FOSS *must* mean Linux, and if anything, FOSS is about having more competition, not less. But they didn't do it in a honest and fair way by creating tough competition, they did it by embracing MS tactics and funnelling a war chest to SCO (the ONLY significant company besides MS who ever paid SCO, aside from fairly negligible entities such as EV1Servers).

A car engine isn't usable without being made, and it should be the same with an MP3 codec: you can patent an translation machine, or a audio recorder/player (and there are such patents), you can't patent any abstract algorithms within it. You can copyright a math book full of formulas or a thesis paper, and you can copyright code in the same way. I think your logic is severely flawed, abstract ideas however complex in art or science aren't patented, only copyright. The patent system didn't serve it's purpose for MP3/Vorbis because it stopped innovation until it could be bypassed altogether.

I agree with that point, but disagree about the definition of "being made". Obviously, patenting the *idea* of a translation machine is nonsensical, but I contend that "making" such a machine consists in devising an algorithm that performs the translation. Coding it (e.g. rewriting the algorithm in an actual programming language) is a comparatively trivial task whose added value is very small, just like patenting a car engine means patenting its internal structure, mechanisms, the geometry of the crankshaft, shape of cylinders etc., not the actual piece of metal that comes out of the assembly line.