IBM's Eclipse OpenJ9 Is A Promising Open-Source JVM

[Ziple]

So Oracle has the Graal VM (a weird JVM implemented in java on which you can execute javascript, ruby, and llvm ir for some reason) as an official openjdk project (http://openjdk.java.net/projects/graal/) well placed to replace hotspot in the open jdk. Now IBM wants to do the same.
That's funny.

[mstoodle]

And why didn't IBM do this 10 years ago?
Why did they wait for Microsoft to open source .NET Core?

The world was a very different place 10 years ago. Arguments that look sensible today or even 5 years ago may not have looked sensible a decade earlier. I have been working on this project for 4 years (it involved a massive amount of refactoring to split J9 into two parts: Eclipse OMR and Eclipse OpenJ9), which probably means the argument was even longer in the making. Just because some things happen near the same time doesn't mean they're necessarily correlated .

[Hi-Angel]

And why didn't IBM do this 10 years ago?
Why did they wait for Microsoft to open source .NET Core?

I absolutely don't see how does it compares. There is for 18 years HotSpot JVM, which is used in OpenJDK.

.NET have won over Java (if that's what you were meaning) due to α) marketing, γ) ecosystem within Microsoft OSes, and β) the C# language being more open to new features. Here I mean β to be a subset of α, because .NET is a buzzword known even to non-technical users — they see it in updates, requirements, some dialog windows. Then, with time, as some of them become "more technical", they go ask/query what .NET is anyway, and finding about C# language.

GNU/Linux do have buzzwords that could be working in a similar manner: GTK and Qt (can't say anything about EFL).

[L_A_G]

This is absolutely beautiful in my mind. Think about it an entire development system all managed under one roof.

Well they say that beauty is in the eye of the beholder so differences of opinion about this are probably to be expected. The way I see this is that it's just going to be a load of duplication of effort to maintain two only semi-compatible versions of the same language. Trying to create something that's fully compatible with both is just going to be a complete mess and thus not a realistic option, meaning that the only sensible option is to kill one and implementing it's best features in the other.

I'm not sure what you are but it might help to see the positive in this.

The only real positive I can see is that the Eclipse Foundation may actually be brave enough to kill one of these and thus significantly reducing redundancy in the Java space and improving one of the two big Java versions. However I don't think Oracle will accept JavaEE being killed of (despite this being the sensible thing to do) nor will IBM accept OpenJ being killed off (which doesn't make much sense when it's instead of killing JavaEE).

IBM already open sourced part of J9 as the Eclipse OMR project last year (see https://github.com/eclipse/omr) and continues to work directly there (you'll see we're averaging or 20 commits per week for most of 2017). Eclipse OpenJ9 will be no different. We're working directly in the open source project as much as possible. We're certainly not "dumping" anything or expecting open source developers to do our work for us. But you're welcome to come join us if you're interested!

I don't think you really understood what I was saying... The transition to a project where IBM no longer contributes isn't obviously going to happen overnight. It's obviously going to be a process that happens gradually where the more outside contributions are made, the fewer internal contributions need to be made and developers can be moved to other projects or just plain laid off (which from what I've seen IBM has been doing a lot of over the last few years).

I'm pretty sure any contributions you'd make would be reduced pretty heavily if you'd have to do them on your own time without compensation...

[starshipeleven]

I hope I'm not the only one who gets the distinct feeling that IBM and Oracle are just shedding parts of their business that don't generate any direct revenue by moving their development over to the open source community so that they can just keep selling their products that generate direct revenue without having to spend any money on developing any of underlying software?

That's what (permissive) open source is all about.
They share the development costs of the infrastructure with others. They give up power over the main project in exchange of money savings. This isn't bad per-se.

[starshipeleven]

I don't think you really understood what I was saying... The transition to a project where IBM no longer contributes isn't obviously going to happen overnight. It's obviously going to be a process that happens gradually where the more outside contributions are made, the fewer internal contributions need to be made and developers can be moved to other projects or just plain laid off (which from what I've seen IBM has been doing a lot of over the last few years).

If IBM decides to drop a project it's still better for it to die as opensource than behind closed doors.
If it dies as opensource it can be scavenged and useful parts can still be ported to other projects, or at least looked at to learn lessons or whatever.

[timofonic]

Everybody's free to develop a JVM. It's getting TCK approval that's the problem and Idk how OpenJ9 manages that.

As far as offloading projects, that's not necessarily a bad thing, since companies fostering them have always been criticized for various aspects and moving widespread products to a more neutral body looks like a step in the right direction. But this leaves open the problem of allocating resources.

I see it a good thing most of the time except when it's covered outsourcing (Apache Foundation is quite known at this). Despite many critics of it (that I agree on many of them I'm able to understand), C++ is very widespread because it's an Open Standard. I disagree the way of classic committees and prefer the lots more productive approach taken by Khronos Group with Vulkan, despite it's a non-profit entity but really managed by big corporations and lacks Open Governing with more community involvement (I think FOSS projects should have representation and vote in relevant ways and not just one corporate like Red Hat, for example).

Java Community Process looks "good", but lacks a more plural and iterative process.

People are afraid of Oracle ways, many forks such as LibreOffice and MariaDB happened because of it too. There's legal actions such as Oracle America, Inc. v. Google, Inc. that doesn't help to trust Oracle too. Some people are considering Oracle to be a patent troll.

I believe if Java wants to survive, it needs to change a lot and become more similar to C++ (and Vulkan) in terms of standardization and legal security of anyone using it. I think it's necessary to make stuff like TCK free and Free Source, but certification to be free and transparent (and better powered by CI infrastructures) if there's no profit.

VERY IMPORTANT: Eclipse Foundation projects use the Eclipse Public License that unlike Apache License 2.0, it's incompatible with the GNU General Public License. This one is a very major issue and reminds me of CDDL.

IBM already open sourced part of J9 as the Eclipse OMR project last year (see https://github.com/eclipse/omr) and continues to work directly there (you'll see we're averaging or 20 commits per week for most of 2017). Eclipse OpenJ9 will be no different. We're working directly in the open source project as much as possible. We're certainly not "dumping" anything or expecting open source developers to do our work for us. But you're welcome to come join us if you're interested!

Those are nice words, but I think it's not enough. What about OpenJDK? What about IcedTea?

What happened to Apache Harmony? There's too many too bad precedents in Java history and make them less relevant will require massive efforts.

Yay, now the open source zealots can prove how awesome their contributions are compared to that of the big evil companies.

Yay, now the proprietary software zealots can prove how selfish and irrational they are compared to that of the big evil freeloaders.

Yeah, that too. Though in this particular case, open source developer are already clustered around OpenJDK. Unless this implementation bring some awesome features and/or killer performance to the table, it will not get traction.

Is OpenJDK relevant really? I mean it may be relevant in Java developers, but most average people install Oracle JRE/JDK when they see "install Java" in their machines.

I install both, because I don't trust in Java "standards". I even install Java on different Operating Systems behind a VM, because I don't trust most Java apps and how they behave in different platforms. Most of the time, lots of Java software work a lot better in Microsoft platforms than everything else.

I suffer lots of crappy Java software required by Electronic Administration from my government (see here as a small but annoying example) required by many indispensable bureaucratic stuff, instead making an ISO standard deployed by as much as more related projects as possible.

The proliferation of Java sucky software might not be a fault of Java itself, but maybe it stimulates it in certain way.

And why didn't IBM do this 10 years ago?
Why did they wait for Microsoft to open source .NET Core?

Microsoft are pirates in suits that became very corporative and big and are use clever parasitic strategies. IBM born as corporation since early days. Despite I dislike any corporation as it gives too much power to one very powerful structure, Microsoft evolved a bit more than IBM even with many unethical & toxic & criminal methods. IBM tried diversification but finally abandoned their IBM PC and moved to big iron, they can't deny their roots and always back to them.

.NET open sourcing isn't real. They just released a bit of it, not the full implementation. They need to play dirty to still stay relevant and still perpetuate Visual Studio as the sacred tool by too many developers it already is.

So Oracle has the Graal VM (a weird JVM implemented in java on which you can execute javascript, ruby, and llvm ir for some reason) as an official openjdk project (http://openjdk.java.net/projects/graal/) well placed to replace hotspot in the open jdk. Now IBM wants to do the same.
That's funny.

Why don't they merge efforts to a common project? Why do they compete each other and divide resources instead making Java suck less?

What happened to Apache Harmony (second mention?

The world was a very different place 10 years ago. Arguments that look sensible today or even 5 years ago may not have looked sensible a decade earlier. I have been working on this project for 4 years (it involved a massive amount of refactoring to split J9 into two parts: Eclipse OMR and Eclipse OpenJ9), which probably means the argument was even longer in the making. Just because some things happen near the same time doesn't mean they're necessarily correlated .

You can't underrate ego in this matter too.

I absolutely don't see how does it compares. There is for 18 years HotSpot JVM, which is used in OpenJDK.

.NET have won over Java (if that's what you were meaning) due to α) marketing, γ) ecosystem within Microsoft OSes, and β) the C# language being more open to new features. Here I mean β to be a subset of α, because .NET is a buzzword known even to non-technical users — they see it in updates, requirements, some dialog windows. Then, with time, as some of them become "more technical", they go ask/query what .NET is anyway, and finding about C# language.

.NET have won over Java, but that's a very relative success. The sector where .NET is popular is the same as Java: Corporate code monkeys.

.NET is a Java clone with some improvements. Common Language Runtime looks a nice concept, reminds me of Parrot Virtual Machine (and interesting project that lacked probably better management and seems mostly dead these days). Despite of that, .NET isn't a 100% ISO standard and is corporately owned, that scares many away just like Java does.

GNU/Linux do have buzzwords that could be working in a similar manner: GTK and Qt (can't say anything about EFL).

[/QUOTE]

GTK and Qt to be similar buzzwords? It depends on who you ask about it. I consider both to be a cancer for the FOSS ecosystem.

On the other hand, EFL has many interesting things on it and still seems to be a capable framework despite it's optimized for embedded systems and multiplatform.

Well they say that beauty is in the eye of the beholder so differences of opinion about this are probably to be expected. The way I see this is that it's just going to be a load of duplication of effort to maintain two only semi-compatible versions of the same language. Trying to create something that's fully compatible with both is just going to be a complete mess and thus not a realistic option, meaning that the only sensible option is to kill one and implementing it's best features in the other.

About killing the best, that's a simplistic approach but it might work. Projects should be looked more as tiny modules joined to make a better one, not as a monolithic product.

They need to merge efforts if they want to survive, that's a reality. Will they success? I have my doubts. Oracle and IBM might get a lot more competitive each other, but IBM is a living fossil that might survive to worse catastrophes than Oracle already suffered.

The only real positive I can see is that the Eclipse Foundation may actually be brave enough to kill one of these and thus significantly reducing redundancy in the Java space and improving one of the two big Java versions. However I don't think Oracle will accept JavaEE being killed of (despite this being the sensible thing to do) nor will IBM accept OpenJ being killed off (which doesn't make much sense when it's instead of killing JavaEE).

Eclipse Foundation is part of IBM tentacles, I don't see how you can see Eclipse Foundation as a neutral entity.

I don't think you really understood what I was saying... The transition to a project where IBM no longer contributes isn't obviously going to happen overnight. It's obviously going to be a process that happens gradually where the more outside contributions are made, the fewer internal contributions need to be made and developers can be moved to other projects or just plain laid off (which from what I've seen IBM has been doing a lot of over the last few years).

I see it just as a covered way to massively deploy outsourcing.

I'm pretty sure any contributions you'd make would be reduced pretty heavily if you'd have to do them on your own time without compensation...

Of course, money buys motivation and not needing an extra activity for getting incomes makes tasks easier in certain ways. But I think there's need to accept and motivate volunteers to participate in FOSS projects, because these newcomers might become less resistant to new ways and ideas than older project members.

[L_A_G]

If IBM decides to drop a project it's still better for it to die as opensource than behind closed doors.
If it dies as opensource it can be scavenged and useful parts can still be ported to other projects, or at least looked at to learn lessons or whatever.

The issue I had with this is that I get the feeling IBM just wants to get other people to continue developing and maintaining this while they make money on selling products that rely on this now community maintained project. That way they get to sell products based on technology developed and maintained by other people rather than selling products based on their own technology.

What it boils down to is really just improving their profit margins by outsourcing work that's necessary, but doesn't generate any direct revenue, to other people.

[starshipeleven]

The issue I had with this is that I get the feeling IBM just wants to get other people to continue developing and maintaining this while they make money on selling products that rely on this now community maintained project. That way they get to sell products based on technology developed and maintained by other people rather than selling products based on their own technology.

And you really think they are so dumb to think that by opensourcing it then platoons of highly skilled people will work for free for them?
Only those that have an interest in it (usually $$$, so companies) will do.

They will probably see if the interest from other companies is there or not, if it's there they will develop it together, if not it will be abandoned.

[L_A_G]

And you really think they are so dumb to think that by opensourcing it then platoons of highly skilled people will work for free for them?
Only those that have an interest in it (usually $$$, so companies) will do.

The move to community maintenance and development is obviously not going to happen overnight, but there's plenty of examples of companies being able to reduce costs by moving to open source software. Just look at how HP killed off HP-UX in favor of moving to Linux and how both Oracle and IBM are in the process of doing the same thing.

They will probably see if the interest from other companies is there or not, if it's there they will develop it together, if not it will be abandoned.

IBM has way too many peripheral products generating actual direct revenue to let this be abandoned. This is either going to lead to them spending less and less money and manpower on developing this and if there aren't any suckers dumb enough to put in the time and effort to develop and maintain the tech for them they're just going to pull it all back in-house with a closed source license.