4 ms·
IMO this is what "source in the preferred form for making modifications" is all about. GNU themselves give the example that C source generated from a perl scrip
by lmm 4y ago
IMO this is what "source in the preferred form for making modifications" is all about. GNU themselves give the example that C source generated from a perl script isn't good enough - if the way you change the program yourself is by changing the perl script, then you need to release that perl script.
I don't think it's reasonable to expect an open-source maintainer to create a better development experience for you than the one they use themselves. If you're not up to editing Whitney-style code in a text editor, you should get good.
But conversely, just publishing a "source code dump", when you yourself work in a much richer development environment, is not good enough. If there's a "how to set up your dev environment" wiki page that you point other developers to, that should be part of your open-source release. IDE project definitions, project structure guides - whatever you use for making modifications to the project is what you need to publish. And if as a maintainer you don't make those steps documented and reproducible, well, that's ok, but in the long run you're only hurting yourself.
In particular, e.g. RedHat's policy of publishing only the "final" source code for their patched kernels, and not the git repo with history (which is what they use when developing it internally) is a straight-up GPL violation as far as I'm concerned.
- zen21 4y ago> I don't think it's reasonable to expect an open-source maintainer to create a better development experience for you than the one they use themselves. This is clearly true. I didn’t read him as demanding that. I think his point is that the fact that something is FOSS doesn’t alone imply that anyone else can actually do anything with it, so people should either not make that claim, or acknowledge what else is needed.
- worthless-trash 4y ago> is a straight-up GPL violation as far as I'm concerned. How ? Are these repos insufficient ? https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-8 https://gitlab.com/redhat/centos-stream/src/kernel/centos-st... https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9 https://gitlab.com/redhat/centos-stream/src/kernel/centos-st...
- lmm 4y agoI was thinking of https://lwn.net/Articles/430098/ https://lwn.net/Articles/430098/ - I remember there being quite a fuss about it at the time, but RedHat employees were clear that it was their company's deliberate policy. Perhaps things have changed since then.
- worthless-trash 4y agoThere has been quite a bit of change in that area since 2011 and I can't fault you for not keeping tabs on it. Bad news always travels quicker than good. Things change more frequently behind the scenes than what appears publicly.
- bdw5204 4y ago> In particular, e.g. RedHat's policy of publishing only the "final" source code for their patched kernels, and not the git repo with history (which is what they use when developing it internally) is a straight-up GPL violation as far as I'm concerned. I've worked in the past on an open source project that is forked from an upstream codebase where the git history and all new tests in recent years are proprietary so I'm definitely very sympathetic to that argument. Its very hard to get people who have no experience in that kind of environment to understand just how much work goes into merging such changes and reverse engineering the bugs from the code changes in order to test for them. In that kind of situation, I think you're better off just hard forking but still keeping an eye on upstream's release notes in case they make a specific change that would be useful to your fork. Trying to maintain full compatibility with upstream releases that don't share their git history is a Sisyphean task and that's probably the intent of many of the corporations that do open source that way.
- chris_wot 4y agoThe other problem is ancient codebases that have been through multiple version control systems. Or started with no VCS at all. For example, any history before 2000 is lost to LibreOffice. What is worse is that for some time the internal processes of Star Division meant that someone bundled up a bunch of code into a single commit, with a terse message about what they fixed. It's very, very hard to tell what code got changed for what issue in those commits. Luckily, this didn't last all that long, but dear god there have been days I've chased back code in the git history to find out why it is there and found it was already there in the original import...
- carterschonwald 4y agoIs this zfs? :)
- hotpotamus 4y agoI thought it was fairly well known that this is intentional obfuscation on Red Hat's part in order to make it harder for Oracle to use their work. I can't blame them on that account.
- lmm 4y ago> I thought it was fairly well known that this is intentional obfuscation on Red Hat's part in order to make it harder for Oracle to use their work. Yes, I know, but that's immaterial to the question of whether it's a GPL violation, or whether it's a moral way to develop software. Free Software means everyone gets the freedoms, even Oracle.
- pabs3 4y agoThe article is mostly talking about things above the source code level though, including the economic and time investment needed to compile/install software, the sheer size of large projects, plugin/etc interfaces for easy extensibility, source code written in unreadable languages (b and APL are mentioned), available programmer time etc.
- lmm 4y ago> The article is mostly talking about things above the source code level though, including the economic and time investment needed to compile/install software, the sheer size of large projects, plugin/etc interfaces for easy extensibility, A lot of that is made easier if the maintainer takes the time to publish all the ancillary pieces they use when making modifications to the project - secondary documentation, build environment setup support, etc.. Yes, there is always a certain level of inherent difficulty in making changes to software - apart from anything else, not everyone is capable of programming at all. But making sure you're really on a level playing field with the original maintainers in terms of tooling, environment setup, and available documentation, makes a huge difference.
- shadowofneptune 4y ago> In particular, e.g. RedHat's policy of publishing only the "final" source code for their patched kernels, and not the git repo with history (which is what they use when developing it internally) is a straight-up GPL violation as far as I'm concerned. I can see what you mean. At the same time, there's always been room for the 'cathedral' form of free software. GNU itself was once much more closed-off when it came to source releases.
- not2b 4y agoThe difficulty with the argument that only publishing "final" versions is a GPL violation is that the FSF itself used to restrict access to development versions of their code and only provide final releases, back in the mid 1990s. Doing development in the open was something egcs pioneered; in the old days only a small set of contributors would have access to pre-release versions and they were strongly pressured not to share it. So it's hard to argue that if Red Hat does things the same way the FSF did for at least its first decade, it's violating the GPL.
- lmm 4y agoThe FSF held the copyright on those projects (and required copyright assignment from outside contributors) didn't they? Whereas although RedHat is a significant contributor to Linux, there are many other copyright holders. (Also, the idea of using source history during development was a lot less widespread then; I don't know if CVS even has a "blame" function, and it would presumably have required connecting to the server if it did)
- not2b 4y agoNot sure what difference this makes. The GPL is binding at the point where someone ships a release. They aren't obligated to ship the details of every internal partial version, which might even include things that can't be legally shipped (confidential information, for example), which is what a complete git history would contain.
- 4y ago
- chrismcb 4y agoWhy? Why do you think then but releasing the git repo with history is a GPL violation? What terms of the GPL are they violating?
- bni 4y agoRMS have stated in several interviews that doing development in the open "bazaar style" is not required. Cathedral is equally valid, and he doesn't care. It's the open source marketing campaign that said that Bazaar is preferred and claim it give superior outcomes. Ive never seen any requirements for documentation. I guess in the 80's 90's code bases were not yet so complex and had som many dependencies as today. I think this article makes a very valid point.