7 ms·
Red Hat ported OpenJDK to 64-bit ARM: A community history
- andrewshadura 6y agoImagine a Red Hat evil twin, developing an Arm64 port behind the closed doors, releasing it under SSPL, requiring a CLA and not accepting merge requests for a potentially paid feature, Windows on Arm64 support.
- geodel 6y agoYeah imagine of the evil of asking money from customers for work done or service provided. Never heard such thing in human history.
- andrewshadura 6y agoNo, the evil of pretending to be open source and pro community, but only really caring about $£€.
- pjmlp 6y agoIf the community chooses MIT/BSD inspired licenses, they don't care about such companies coming into action.
- josephcsible 6y agoHow often does the community choose a non-copyleft license? What I always see is a company choosing a non-copyleft license.
- spijdar 6y agoI'm a sample size of one, but I often pick a permissive BSD/MIT license for the personal software I develop, and at times will pick BSD-licensed over GPL-licensed software. I prefer the brevity and clarity of the shorter licenses, personally. I've also found that, sometimes, if there are two similar projects with a BSD vs GPL split, the BSD project tends to be more pragmatic and focused on the product/code more, while GPL projects err towards ideology, possibly at the cost of functionality. OpenSSH vs lsh as an example.
- pjmlp 6y agoAll the time, in the OpenBSD, NetBSD, FreeBSD ecosystems. In fact, being anti-GPL has become kind of fashionable, so there you have it.
- ryukafalz 6y agoThankfully they couldn't; OpenJDK is released under the GPL. Copyleft works, people!
- Alupis 6y ago> Thankfully they couldn't; OpenJDK is released under the GPL. Copyleft works, people! Not quite. There are many commercial JDK/JVM offerings that are closed source and survive very well with their "secret sauce" (usually GC related), such as Azul's Zing JVM. Red Hat's customer base that used JavaEE containers (and just Java code in general) was large enough that Red Hat could have easily decided to implement and maintain their own JDK offering - but they didn't. Probably based on Red Hat's history of being a good open source "netizen", they chose to use the GPL code even though they did not have to. We're all better off because of Red Hat's contributions to OpenJDK (porting to 64 bit ARM included) - but that's not exactly "Copyleft works!".
- ece 6y agoRedhat is a big contributor to OpenJDK, so there's really less of a "maybe not" than you think. Even Azul maintains an OpenJDK distribution, so copyleft is playing a part in keeping things more open and supported than not.
- Alupis 6y agoThe part OP missed, and perhaps you may be overlooking, is none of these companies are required to use OpenJDK (and the GPL'ed code) for some unavoidable business reason. They opted-into using it consciously. Red Hat because they embrace Open Source above all else (or at least did at the time), and Azul as opportunists during the decline of Oracle JDK as a way to get Java users into their ecosystem (and eventually make money). OpenJDK is wildly popular as the defacto reference implementation of Java, due to moves Oracle made a few years back (on purpose to push Java language development into a more community-driven approach). OpenJDK is good, but there are reasons people seek out custom implementations (massive heaps, pauseless GC, better tooling and monitoring, re-implemented IO, etc). All that is to say, contrary to OP's original assertion, it's not that Red Hat "couldn't" release some closed source JDK... they chose to embrace the GPL'ed code. The notion that "Copyleft Works!" is by compelling companies to open source their GPL'ed code. That's not what happened here.
- INTPenis 6y agoThat's business as usual. Open source is breaking the norm of the corporate world and that is usually what makes a product like raspberry pi gain traction. Because it's accessible enough for regular people to play with for free. Countless manufacturers have made closed products and sold them for license money. Nothing new there.
- smspf 6y agoThis is nice and makes Red Hat look like an amazing OSS contributor, which I think they are. But truth be told, things were not that easy when it comes to RHEL and derivates on aarch64. As contractors working on getting stuff working on aarch64, I know we struggled quite a bit to get CentOS7 to get to the level Ubuntu was at the time on aarch64. And real progress (i.e. getting a fully working distro for our intents at the time) was only seen after money were thrown and a contract was signed. The bottleneck however paid off imo - Red Hat insisted on doing things slow and steady, so once things started landing in the distro, they just worked. Unlike other distros which touted aarch64 support early, but struggled to fix issues along the way. I remember working with OpenJDK and Opendaylight on aarch64. The crashes I've seen were terrifying. The JVM crashed so hard, our expert Java developer laughed 30 minutes straight looking at one trace. To this day I have no idea what was so funny, as he couldn't explain it in a simple manner and I lost interest. Anyway, thank you Red Hat developers for all the work (minus the systemd "architects", but that's a different story).
- bitcharmer 6y agoJava for ARM is really exciting. I'm wondering however how Java Memory Model implementation performs on ARM. JMM offers pretty strong guarantees via the fundamental happens-before relationships in program behaviour. It's relatively easy to implement on x86's strong TSO, but ARM is extremely problematic to offer the same guarantees on. Can't wait to run some concurrency benchmarks on this platform.
- monocasa 6y agoPowerPC has pretty much the same memory model as the ARM variants, so it's a more or less solved problem thankfully.
- dralley 6y agoOn that subject, as someone that knows nothing about either POWER (or Aarch64), I'm curious if anyone could describe the ways in which they are different?
- aph_at_redhat 6y agoThe most important practical difference is that Power PC is not multi-copy atomic. That is to say, "[the PPC] memory system does not guarantee that a write becomes visible to all other hardware threads at the same time point." AArch64 does guarantee this property, although it was not documented as doing so in the first release of its specification. For this reason PPC has to emit some extra StoreLoad memory fences. (See "A Tutorial Introduction to the ARM and POWER Relaxed Memory Models" by Luc Maranget, Susmit Sarkar and Peter Sewell, but be aware that the AArch64 specification has changed since that paper was written. Our work has resulted in Arm clarifying the AArch64 specification in a couple of places, for which I'm grateful.)
- aph_at_redhat 6y agoWell, maybe PPC doesn't strictly have to do this, but the PPC maintainers decided it would be surprising if it didn't.
- 6y ago
- BenoitP 6y agoNice! I guess the next target is to do RISC-V. I think it only works with the interpreter at the moment.
- aph_at_redhat 6y agoIt's much further along than that. See https://github.com/isrc-cas/bishengjdk-11-mirror https://github.com/isrc-cas/bishengjdk-11-mirror
- nerpderp82 6y agoI love this writeup, esp around how the dev team architected the work to derisk the work before hardware was available and how the emulator/instruction interpreter actually sped up development velocity over using hardware directly.
- aph_at_redhat 6y agoIt worked far better than I expected. Around the turn of this century I was involved in bringing up the software for a new processor made by a certain large semiconductor company (you might be able to guess!) The early hardware didn't work perfectly, the operating system wasn't ready, and the compilers were flaky. The cost of writing a simulator was cheap when compared with avoiding all of that pain and getting first-mover advantage.
- ossusermivami 6y agoNo matter what happened with CentOS, i have huge admiration to redhat, the IT industry would be a very different world if not for Redhat driving a lot of the OSS
- compsciphd 6y agoas a totally tangentially related thing, we just released a proof of concept version of kodi with java on aarch64 for android. i.e. first time in the open source world one can play blurays on a 64 bit arm android os with full java menus (well, that's mostly the nvidia shield pro 2019, but should be extensible to 32 bit android as well). https://forum.kodi.tv/showthread.php?tid=360250 https://forum.kodi.tv/showthread.php?tid=360250