4 ms·
When this came up yesterday[1] I made a comment that a lot of people seemed to find interesting (by votes, not responses: there's nothing in it really to respon
by saurik 13y ago
When this came up yesterday[1] I made a comment that a lot of people seemed to find interesting (by votes, not responses: there's nothing in it really to respond to), so I figure I'll make it again today.
[1]: https://news.ycombinator.com/item?id=7116144 https://news.ycombinator.com/item?id=7116144
In case anyone doesn't realize this: the ARM64 LLVM backend used by Apple in Xcode 5 to target, for example, the iPhone 5S, is currently not open source. There is a vague promise that it will be merged, but the time frame described was itself long, many months have already passed, and it doesn't really seem like much progress has been made since (although of course things could be happening internally at Apple).
(In fact, none of the code actually backing Xcode 5 or iOS 7, even LGPL code such as WebCore/JavaScriptCore, has yet to be released; but, in the case of the ARM64 backend, there was a specific discussion about it on the mailing list, and in fact some iOS-specific ARM64 patches to LLVM languished seemingly due to the idea that "well, some day the ARM64 stuff will be merged from Apple", so it is already having "chilling effects".)
- BudVVeezer 13y agoA quick search for ARMv8 and AArch64 in the LLVM commits list shows that a considerable amount of work has been happening on this front. In fact: http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20130107/161182.html http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-... I would have a hard time saying there's "chilling effects" happening.
- saurik 13y agoLLVM has their own ARMv8/AArch64 backend that is unrelated to Apple's. This backend, as far as I understand, does not target Apple's modified conventions and file formats for iOS. A patch was provided for this, and as far as I understand it was not merged. The patch was ignored on the mailing list, and even came with the caveat attached "I'm not sure if the llvm maintainers want the patch given the previous message that there's going to be an official patch set from apple to support this, but here is mine." http://lists.cs.uiuc.edu/pipermail/llvmdev/2013-September/065936.html http://lists.cs.uiuc.edu/pipermail/llvmdev/2013-September/06... For further evidence of this, we can look to Haskell: Haskell currently has been unable to target iOS's ARM64 architecture because of this part of the compiler not being open source. It is precisely these kinds of ripple effects that Stallman is concerned with: he would argue that by supporting LLVM the community set themselves up for a situation where Apple was able to later hoard valuable changes while still getting the benefit of an army of people fixing bugs in their compiler for free. > Also in "Build Settings", change "Architectures" to "Standard Architectures (armv7, armv7s)" (as opposed to "(including 64-bit)") as Apple hasn't yet merged Arm64 support into the LLVM source that we'd need to fully support 64bit iOS devices. The code will still run wonderfully on the 5S/iPad Air/mini Retina in 32bit mode. https://ghc.haskell.org/trac/ghc/wiki/Building/CrossCompiling/iOS https://ghc.haskell.org/trac/ghc/wiki/Building/CrossCompilin... Honestly, it is as if you just saw "ARM64", did a Google search, took the first hit, and decided "that's enough research for today"... :(. Your citation is from over a year ago, long before all of this happened, and is not relevant. I specifically stated in my comment that the issue was with relation to the "ARM64" backend as it targets "iOS", not the general "AArch64" backend as it could be used to target other platforms or operating systems. This is a critical distinction when talking about Apple.
- BudVVeezer 13y agoI'll ignore your ad hominem and simply point out that Tim (the person who commit the AArch64 backend from the post I linked) is an Apple employee...
- saurik 13y ago> Tim (the person who commit the AArch64 backend from the post I linked) is an Apple employee At the time of this patch, Tim worked for ARM, not Apple. Tim was hired by Apple sometime during 2013, as far as I understand after he was working on committing these patches. This information is thereby also not relevant :/. (If anything, it is worrisome.)
- wonderingg 13y agoBut if you ignore ad hominem and other fallacies you have to throw out 90% of what he says in most comment threads.
- the_mitsuhiko 13y ago> It is precisely these kinds of ripple effects that Stallman is concerned with: he would argue that by supporting LLVM the community set themselves up for a situation where Apple was able to later hoard valuable changes while still getting the benefit of an army of people fixing bugs in their compiler for free. In the absence of LLVM apple would license/develop a closed source compiler. They would not use a GPLv3 GCC if they had to.
- thatthatis 13y agoWith copy middle there are always going to be examples of "x wasn't re released into the community." I think focusing on these is extremely short sighted. It cedes a premise that the way to judge the health of an open source paradigm is %_advances_open_sourced, or as a negative indicator #_advances_not_open_sourced. The true equation for the value of an open source paradigm should be: #_advances_open_sourced, and nothing else. It doesnt matter if 99% of advances stay closed, what matters is if there is enough value repatriated. Copy middle allows a company to take something open source, invest speculative resources on improving it, and it successfull sell for a portion of the incremental value - both improving the quality and decreasing the cost of software to consumers. Then, if/when a company tries this and fails to find a market, the advances sometimes/often get repatriated as part of the shutdown. Copy left tries to use anti-ownership ideals to ensure software remains free. This seems like a good idea, until you realize that copy middle effectively harnesses capitalist motives to make free software better. In my mind: Copy left is, by design, a virus. Copy middle is a symbiotic guest organism. I know when given the choice between a probiotic and the flu virus which I want to ingest into my body. I have the exact same preferences when it comes to what I invest in my company.
- saurik 13y agoThis is your opinion. I will not even attempt to say it is not a valid opinion: a lot of people I know hold this opinion, and at least in the short term it seems to be a consistent opinion. I think there is an interesting discussion to be had "in the long term", but it is just that: a discussion. However, I am going to assert that it is, in fact, an opinion. The thing you need to understand is that the other side of this particular opinion does not actually care about either "#_advances_open_sourced" or "%_advances_open_sourced": the goal is freedom, not openness, and these are assets that are neither equivalent nor convertible. To put this into a very simple example: if Apple open sourced every single software component on the iPhone, that would be extremely "open". We would know everything about how the iPhone worked, and we'd have learned a lot as a community about how to build mobile phone operating systems. This would help other people and companies work on mobile phones, and might have positive effects on things entirely unrelated to mobile phones. To the people who care about variables related to "#_advances_open_sourced", this would be a great outcome, and "%_advances_open_sourced" is even 100%: seems great. But, in such a universe, I will argue that the users of an iPhone are still not "free" because they are unable to take the software running on their iPhone and replace it with modified software based on all of that wonderful open source code. In fact, lest you say "people then shouldn't buy an iPhone", it might be the case that users of any device are not free, because there is absolutely no guarantee in such a universe that any of the numerous devices that are based on this open source software--even if 100% of the software on these devices is itself open source--actually allow users to modify the software running on them (hence, GPL3). This is the difference between "free" and "open", and in my understanding this is why Stallman makes the points he makes: he doesn't care whether there are more cool and fun toys built out of open components, what he cares about is whether the toys we have restrict our freedom. It might be that a world where we are more free has fewer toys. In my interpretation, Stallman would consider that a "win" for mankind. In his mind, that argument "so don't buy an iPhone" seems to be related to the argument "so don't link against their code", and he wants people to stop: the idea is that if you don't like people restricting freedom, you should actively avoid yourself supporting it. (So, that's all I really have to say in response. However, if you--or anyone else--finds this kind of stuff interesting, I spent a lot of time a few months ago writing a lot about this subject, and in particular this tradeoff, in a comment thread on Hacker News when FreeBSD deprecated gcc. Of particular highlight, I go into detail on the property that GPL is more a license about binaries and hackability as it applies to end users than about source code and extensibility as it applies to developers, as well as why this causes debates about this subject among a bunch of developers to be problematic, due to the different incentives developers have from end users.) https://news.ycombinator.com/item?id=6209724 https://news.ycombinator.com/item?id=6209724