IBM Offers to Help Sun Open Up Java

dave writes "ESR has opened the issue of pressuring Sun to open source Java, and today IBM throws in their own commitment toward this end. IBM has published an open letter to Sun, proposing that the two companies collaborate on an independent project to open source Java, saying that IBM is ready to provide technical resources and code for the open source Java implementation while Sun provides the open source community with Sun materials, including Java specifications, tests and code."

Comments

How nice of IBM..

[grub]

Just playing Devil's Advocate here: IBM sounds touchy-feely about open source but how would they react if Sun were to offer to help IBM open up AIX?

Re:How nice of IBM..

[Anonymous Coward]

I'm not sure that this comparison is totally valid. IBM isn't marketing AIX as a counter-culture product that challenges the dominance of the 'evil monopolist' Microsoft.

Sun is trying to have it both ways- claim their language will "free" them from the clutches of MS while trying to clutch the language specification/certification for its own.

Re:How nice of IBM..

[d00ber]

IBM put JFS the AIX journaling filesystem into Linux. They might well respond by saying yes. I do admit that given IBMs long history of tweaking SUN (Eclipse anyone?) that this offer by IBM might not be taken that well. OTOH Sun has never seemed that hell bent on monetizing Java - the licensing and conformance testing fees probably don't begin to cover development costs.

They say are trying to build a community around Java and it seems to me that given those two points Sun really should release Java to the open source community.

I think this really would give Java a HUGE boost over .NOT.

Besides, JBoss and Linux distros make money on packaging and supporting Free but hugely complex systems.

Re:How nice of IBM..

[fantastic]

Dear IBM

Can you opensource DB2, we don't have a database at Sun so this would really help

Re:How nice of IBM..

[Coz]

Brav-o. Good summary. One little thing - if Sun open-sources their own code, it's not duplicating the API - it's releasing it. Now, if there's IP in that code that's locked up somewhere, or licensed somehow, that's a problem - one that IBM should be able to help solve, given their library of IP rights.

If I were at Sun, one of my concerns would be which of their development projects to open, and when. "Java" isn't just Java 2 Standard Edition - the Enterprise Edition and Mobile/Wireless Edition have lives of their own; then there's (still) Jini and all their XML stuff. Sun is sinking cash into lots of different efforts, trying to establish Java in market niches (like mobile phones) and building in tool support, documentation, etc. Throwing the doors open and letting the world at their code base may not be the smartest thing at the moment (esp. if there's licensed IP in there somewhere that they need to go negotiate to open, or remove).

I'd like to see them phase in open-source. Give 'em six months or so for the 3 major "platforms", including all the java.* and javax.* packages, then another six months for the com.sun.* packages - with an expectation that other players would start working on them immediately. After that, every new thing they do should be opened no later than beta... and the JCP should allow participants to collaborate on implementations at the source-code level, so JCP members could work in semi-privacy until the code got fully opened at their beta release.

But that's just an idea....

Re:How nice of IBM..

[AstroByte]

Come on then, Mr. Big Mouth. When YOU pick up the Virtual Machine spec and implement your own VM from scratch, THEN I'll believe you when you say any monkey can write a VM. In the meantime, talk about things you actually know something about.

I HAVE written a VM from scratch, and worked on Sun and IBM's Virtual Machines. I can tell you from experience that writing a VM is definately not a trivial task. In fact, it is probably harder than the libraries. The libraries are by definition Java code. The major problem is the sheer size of them.

A modern VM on the other-hand, covers a wide range of techniques. Writing an efficient thin-locking implementation is far from trivial - the code is extremely complex, and even a slight mistake can lead to race conditions, leading to unexpected behaviour which is very difficult to track down.

Likewise, a modern garbage collector is an advanced field in itself (e.g. parallel collectors, generational collectors, etc.). Again, a simple mistake can take weeks to find.

Have you also forgotten about the JIT? Or more accurately a DAC (dynamic adaptive compiler). Whereas a standard compiler can spend as long as it likes optimising the code and be slow as hell, a modern VM must profile the code on the fly, and transfer control between compiled and interpreted modes efficiently. Again, not trivial.

Even following the spec is non-trivial. There's enough grey areas to cause a VM implementor to pull their hair out.

Sun and IBM have large teams working on these VM's, many from research backgrounds and with PhDs (including me). Thanks for calling us all monkeys.

Re:How nice of IBM..

[Anonymous Coward]

You're almost right.

Not "any monkey" can write a VM. They're non-trivial (just ask the Kaffe folks), and IBM has several irons in the fire. As well as licensing Sun's VM (and improving it vastly for their customers) they also have their own VM under development, entirely free of Sun IP.

On the issue of class libraries, you're also nearly right. Swing would be hard work, and pointless. There's a reason that eclipse doesn't use Swing... IBM isn't interested in it - it sucks.

IBM also has their own set of class libraries under development - entirely free of Sun IP.

So, in my opinion, this is just a huge red herring. IBM has enough projects under heavy development to release a completely open-source VM and set of class libraries within 18 months if they want to.

Personally, I think it's going to happen, and this is them tapping on Sun's window going "if you don't do it, we will, and we'll do it without using your IP".

Re:How nice of IBM..

[ulrikp]

If Microsoft "borrowed" GPLed code and tried to hide it not only would they open themselves up to a serious lawsuit from the copyright holders (with serious monetary penalties), but they could theoretically end up having to share any source code that came in contact with the GPLed code.

This is what the FUDsters would have us believe, but it's not true. As Eben Moglen, General Counsel for the FSF, has repeatedly pointed out, the GPL is a license, not a contract.

One of the consequence os this is that you can't force someone to open up their own code if they link against GPL'ed code in violation of the license. At most, you can force them to stop using the GPL'ed code.

See this lengthy rebuttal by Pamela Jones of Groklaw fame, or this more accessible, shorter version.

Ulrik

Re:How nice of IBM..

[AstroByte]

I admit by first post was a bit OTT -- it's been a long and hard day today (it's 20:15, and I'll be here a while yet). No, it isn't working on a VM :)

I agree that there's a long way between a basic VM and a state-of-the-art one. However, you were making the point that the libraries were the stumbling block to a truly free Java implementation. My point was that it is the sheer size of the class library that's making this difficult, not the complexity of the code per se.

While it may be possible to write a _very_ basic VM in a couple of months(*), it'll be just as unusable as a half completed swing implementation would be. We need both a complete class library and a state-of-the-art VM. One without the other is pointless.

(*) Having said that, there's been a lot of feature creep in the VM to keep up with the ever expanding APIs. Support for Reflection is fairly non-trivial as is class-loaders. Neither of these are covered in the VM spec (class loaders obliquely in places). Nor is GC support for weak, phantom, etc. references. In fact, you don't need a GC to fulfill the spec. A toy VM will not only be slow, but very incomplete as well.

Why doesn't IBM open source it's own VM?

[GGardner]

IBM has a less well known Java VM for embedded systems called j9. This was developed in a clean-room way. If IBM wants an Open Source, commercial quality VM, there's nothing stopping them from opening this one.

Re:Why doesn't IBM open source it's own VM?

[pragone]

Perhaps they acknowledge that if a standard is desired, they should contribute to a popular JVM rather than keep working on one themselves. I think that's a good way of thinking for OS.

Not said...

[ackthpt]

Hi Rob,

[Open Source Java or you risk relegating it, while .NET on commodity hardware gobbles up both the development and hardware markets to Sun's eventual doom. Work with us and Java will be strong as many eyes and hands (ours included) clean it up and expand it where need and demand lay. Ignore this request and we'll pick it up at your bankruptcy auction.

Regards,
Rod

Nice.

[NegativeK]

This is a win-win situation for IBM. If Sun goes for it, IBM gets it's hand in the development of an open implementation of Java. If Sun doesn't, IBM can say that they tried. Either way, IBM appeals to the open source community.

As for Sun, it's almost lose, lose. If Sun goes for it, they lose complete control over Java, which is a cash crop for them. If they don't, they look stingy to the open source community, and alienate a lot of us.

Poor sun, nice IBM. .

Re:A Question

[pubjames]

What changed, specifically?

IBM are being very intelligent. They are moving with the market.

It used to be that everyone in the IT world was closed and proprietary. OSS is changing that, and IBM know it. IBM are going with the flow, not fighting it.

ESR: not so bad after all?

[Hrolf]

So does this mean the Slashdotters who claimed that ESR damaged the open source community via his letter to Sun will now retract their views? It seems we're closer to an open source Java as a result of his opening salvo (little gun-toting humor there) than we were before he wrote his letter.

Re:ESR: not so bad after all?

[hambonewilkins]

Eric! What have you been up to?!

Sorry, forgot, I'll just check your webpage for info.

Licensing issues

[MAXOMENOS]

Q: What's to stop Sun and IBM from open-sourcing the JDKs they have now? A: Third-party IP. Odds are, both Sun's and IBM's JDKs are chock full of third-party IP. Even the stuff that IBM implements in a "clean room way" probably contains IP that IBM licensed from somebody else. One could interpret IBM's gesture as offering to produce parts of the JDK that are free from IP encumbrances.

Listen to Ganesh Prashad

[IGnatius T Foobar]

Everyone's read ESR's open letters, but the real, convincing, extremely well-written case was done by Ganesh Prashad in a Linux Today editorial yesterday. Ganesh lays it out in terms Sun can understand, without ESR's controversial style. This article is a must read for us, but it's also something that should be absolutely wallpapered in Scott McNealy's office, and maybe his home too.

Ganesh very clearly demonstrates how Sun will lose J2EE's 'lingua franca of business logic' status to .NET if they don't let the community galvanize and help out, and the only way to do that is to open source the Java core.

Whoa...

[bersl2]

Did ESR just bitch, and things actually happened?

I'm impressed; unless he has an "in," of course...

Re:Doubtful...

[Frymaster]

I haven't seen Sun do anything that leads me to believe they are really for open systems.

then perhaps you should take a look at experimentalstuff.com - sun's site for experimental code. lots of it is opensource including an entire operating system (chorus os).

looks like a committment to opensource to me.

Re:Doubtful...

[Anonymous Coward]

*Cough* OpenOffice *Cough*

Re:Doubtful...

[egomaniac]

Does anyone actually think this is actually gonna happen? Sun has always impressed me as a Microsoft wannabe. The only reason they are currently allying themselves with Linux is because "The enemy of my enemy is my friend."

Oh, bullshit.

When Microsoft wants to include a new technology (such as their web browser), they integrate it in such a way that you are forced to use it, can't replace it with something else, and end up using it for unrelated functions whether you want to or not. And half the interface is undocumented and inaccessible to people outside of Microsoft.

Sure, you can download and install your own web browser, but IE is always there, sucking up resources, no matter what you do.

When Sun wants to integrate something new into Java, the JCP develops a standard API for accessing it. Sun develops or licenses their own implementation of it, but allows you to use any implementation you choose so long as it implements the standard. Other VM distributors are free to replace Sun's implementation with their own. In fact, it's remarkably easy to be completely unaware of which implementation you are using.

If Microsoft had developed a standard web browsing API which allowed you to swap out IE in favor of Mozilla, and allowed computer manufacturers the right to do so on preinstalled machines if they chose, nobody would be complaining about the IE integration.

Well, that's exactly how Sun would have chosen to do it, based on their track record. They make it remarkably easy to swap out portions of the runtime library with alternate implementations.

Open Source dangers...

[brasten]

McNealy addressed this issue year or so ago...

The problem they're afraid of is Microsoft embracing & extending Java. The ability that Sun had to sue Microsoft and force them to cease their modifications would no longer exist.

Now imagine Bill Gates at home in his Medina mansion.. (only 10 minutes away from here actually... sad...). Everywhere he tries to push .NET and his vision, Java's there. Java's beating him, or right behind him, on almost every front, and for the better part of the last few years, he's been unable to combat this enemy with any major success. Now imagine someone hands him the source code and tells him he can fork it however it wants. What would he do?

I don't know. And for the time being, I'm fine not knowing...

Re:What about gjc?

[dominator]

IBM could do as you suggest. Then again, they have already written their own Java compiler (Jikes), at least one of their own JVMs, their own servlet container (Jakarta), etc...

I'd recommend looking at this page for more info on IBM + Java + OSS.

IBM has already written at least one high-quality JVM implementation which is not OSS because of contracts that IBM has with Sun. Of course, suggesting that IBM work on GCJ and Classpath has some merit in and of itself. But realize that IBM has sunk untold man-hours and dolars into developing its own JVM - resources that they now wish to contribute to the community at-large as OSS. I personally can't blame them if they didn't wish to spend a similar amount of resources on GCJ and Classpath when what they've got works.

Perhaps with this Open Letter IBM is looking for permission to open up the code. Perhaps they are looking to collaborate with Sun to create an even better project. Perhaps this is all just marketing/PR bs. Time will tell.

Dom

Re:A Question

[Mr. Slippery]

But they are still one of the leading hardware companies

Exactly. I spent some time contracting at IBM, and that's the main thing I learned about their business model: all the software stuff they do has the goal of selling IBM hardware. That's a plan that plays quite well with open source / free software.