Those who hope Sun will open-source all of its software products anytime soon are in for a big disappointment. Sun executives, including president Jonathan Schwartz and John Loiacono, the executive vice president for software, have all repeatedly said that the company intends to open-source its entire software stack over time. However, they have not been specific about the time frame for that, which has left the impression that it is imminent. But Simon Phipps, Sun’s chief open-source officer, said in an interview that this process is not going to be rushed and will not happen quickly.
“But Simon Phipps, Sun’s chief open-source officer, said in an interview that this process is not going to be rushed and will not happen quickly.”
Should it be otherwise?
Sun has been into open source before “open source” existed.
They have a decent track record (Openoffice.org, a multitude of Unix software like nfs, now OpenSolaris, ZFS, etc).
I just wished they’d get it over with and open source Java. Although, it’s their own neck they are cutting. GCJ is coming up faaast.
“Although, it’s their own neck they are cutting. GCJ is coming up faaast.”
Only with a tailwind, going downhill, and a rocket pack strapped to their back.
“I just wished they’d get it over with and open source Java. Although, it’s their own neck they are cutting. GCJ is coming up faaast.”
Well, I wouldn’t say GCJ is coming up that fast. Right now, applications that will run under GCJ are the exception rather than the rule. And it has been in development for quite some time. I have more faith that Harmony will produce a truly usable open source JVM than GCJ will.
Given that gcj is not a JVM, but an ahead-of-time compiler, and does not try to build a JVM, I’d expect any project building a JVM to produce a truly usable JVM faster than gcj. Heck, many have. See JamVM/Cacao/IKVM/SableVM/JikesRVM/*
That doesn’t mean that gcj ain’t very useful: see fedora core’s selection of packages that run and build on gcj. A lot of stuff works, and a lot more will, as the missing and broken bits of the class libraries get fleshed out. J2EE servers are now within reach, with JOnAS being the first J2EE server to be packaged for FC 5, afaict, and performing increasingly as correctly under gcj as it does under proprietary runtime implementations.
Now, with JOnAS, and possibly other J2EE servers humming along on gcj next year[1], a lot of things have to be in place, and working correctly. The dependency graph of something like JOnAS is very, very huge. Given that Java’s strength in the marketplace is on the serverside, I’d hope that gcj can help make serverside Java programming more accessible to people, by making it easy to package the bytecode jar mess as plan old DSOs, and have it come with each distro out of the box, like, say, Ruby, PHP, Python and all those succeful other ‘little man’s frameworks'[2] do.
That, in Sun’s JS’s own worldview means bringing in more volume for the platform, so it is a fine thing for all of us peddling in Java, free or non-free.
Harmony will hopefully produce an interesting, usable VM in a few years, as well, but gcj is already increasingly usable today, and will serve us well until that Harmony VM of the future appears.
None of this has anyting to do with the things Simon said in that interview, anyway, so it seems like we’re getting a bit sidetracked here.
cheers,
dalibor topic
[1] Resin’s new, fast PHP on gcj … now that would be interesting.
[2] That’s how they are often seen from the Java side of the fence, apparently. See the current ‘fear and loathing in Javaland’ regarding Ruby & Rails. Last year it was PHP, and the year before that it was Python, and the year before that it was C#.
”
Well, I wouldn’t say GCJ is coming up that fast. Right now, applications that will run under GCJ are the exception rather than the rule. ”
hmm, see http://fedoraproject.org/wiki/JavaFAQ. Openoffice.org base, Eclipse, Jonas, Tomcat etc run on GCJ with Fedora Core 4 and the numbers keep increasing everyday. Just look at the Fedora development tree. GCJ is about 90% complete with Java 1.4 with a fast paced 1.5 branch. if that isnt open source Java what is?
GCJ is BOTH a compiler as well as JVM (gij) and Fedora Java development from redhat people can switch between being a compiler, gij JVM or Sun JVM dynamically as well as manually
“GCJ is about 90% complete with Java 1.4 with a fast paced 1.5 branch. if that isnt open source Java what is?”
Because that 90% is misleading since it only includes the core Java API. Virtually nothing from javax.* has been completed for example. There is no Swing support, and so on.
well from a companies point of view it makes sense not to jump fully into open source quickly a slow graduale changes is more sensable.
It`s good to wait for great os stuff
Not really sure why this is news.
Not really sure why this is news.
Admitting to vulnerability in this forum isn’t going to get you any dates. (^;
No link back to Maureen O’Gara’s article that I pulled the following quote from. It’s available from other [read reputable] sources, I’m sure.
“We are going to be interested in seeing how they do that because we own Unix. We own the copyrights and the patents,” Novell CEO Jack Messman said during Novell’s Q4 earnings call Thursday. “Solaris is somewhat built on that.”
That was a year ago (Nov. ’04). OpenSolaris now exists in the wild and no further comments have been made by Novell that I am aware of.
Perhaps the news is that Sun has made OS available without publishing any encumbered code and that they are slowly heading that direction with other pieces of their stack.
In case you might think that the current process appears to be slow consider this (from Jan of ’94):
“UG470-01 SUN TO STRIKE OUT ON ITS OWN: BUYS OUT UNIX SOURCE RIGHTS FROM NOVELL
Sun Microsystems Inc was reportedly closeted with Novell Inc last week cutting a deal to buy out the rights to its Unix-based Solaris source code. The purchase, which insiders said could cost Sun anywhere from $90m to $125m, would mean it never again has to pay royalties for its operating system to Unix’s titular owner.
It would also allow Sun to license the code to other Unix vendors, develop the thing in its own way and collect all the royalties for itself. This would effectively elevate Sun to the position of “Second Rome” in opposition to Novell’s tarnished orthodoxy, possibly refragmenting the industry along major new fault lines and driving despairing OEMs into the Microsoft Corp Windows NT fold. The move is a logical one for Sun, which has been dead set on such a course since the middle of last year when its SunSoft unit organised a swat team to proselytse Solaris among Unix OEMs and top PC houses (UX No 453). Its first convert
was Amdahl Corp (UX No 454).
Last week’s deal was said to cover only Unix SVR4-derived code, putting potential customers Hewlett-Packard Co and IBM Corp beyond Sun’s immediate reach unless it can persuade them to the unthinkable – adopting Solaris. HP-UX and AIX are derived from pre-SVR4 Unixes. HP, however, is believed to currently be negotiating the rights to its HP-UX binaries from Novell. Sun is expected to pay Novell about $15m this year in binary royalties. Projecting on the price Sun could be paying – probably around the $100m mark – it could represent five or six year worth of royalties provided its market maintains current levels. Unix System Labs, under either AT&T Co or Novell, never cut such a deal to knowingly set up its own competition. In the only other arrangement that is the least bit similar, Cray Research Inc last year bought out its source code license, but the intention was in no way the same. Its customers need source code and were used to buying it and paying royalties to both USL and Cray. Novell is said to be aware of the risk it is running letting Sun set itself up as an alternative Unix source. However, it is also apparently betting that it can outmarket Sun with its SVR4.2 UnixWare kit.
Currently, it looks a foolish bet since it reportedly sold only 35,000 packages since the thing hit the market 13 months ago. Sun, on the other hand, has been meeting considerable resistance to its Solaris 2.x software, with its users reportedly forming “4.1.3 or Die” clubs. Observers are highly sceptical that the number of Solaris 2.x units SunSoft says it has shipped are actually in use, and the success of its swat team effort has yet to be proven.”
So 12 years between buying up the rights from Novell and releasing OS seems like a long time to me. Maybe they are being cautious instead of wishy washy as some might call them.
“Admitting to vulnerability in this forum isn’t going to get you any dates. (^;”
Admitting to reading it isn’t going to get you a date either, much less admitting to contributing. LMMFAO
I admit nothing. I only imply…
Harmony is currently vaporware (Keep in mind I volunteered to help work on it). GCJ is usable for many things, particularly non-vendor proprietary code. For example, SWT-based applications tend to work great with GCJ because SWT is open source. Swing apps do not, because Swing is closed source and they have to reverse engineer it.
GCJ isn’t really about being a JVM anyway, it’s more or less about using the Java language and the huge amount of code available for it. Although GCJ does have JVM abilities, if you need them.
For most applications, the JVM is the weak link in the chain. The JVM is why people bitch about memory issues, long startup times, etc. The language and development platform itself isn’t the issue (unless you’re a pedant).
“Harmony is currently vaporware (Keep in mind I volunteered to help work on it).”
It is vaporware. But it is Apache. And Apache has a pretty good track record of turning vaporware into reality. Usually when Apache says they are going to do something, they do it, and they get it done. It’s very rare that an Apache project fails, which I suspect is one reason why when Apache sponsors a project, it tends to get support from industry players–they know they won’t just be throwing their investment down the drain on a project that is going to die out in a year or two. Even Sun has said they will probably contribute code to Harmony. And Sun is one the biggest soothsayers when it comes to point out “I don’t think people who want to build their own Java implementation realize what a massive undertaking it is.”
Sun also offers free compatibility and certification testing to recognized open source organizations. So Harmony could even make the goal of trying to be 100% comptible with Java. If they can, Harmony might even be able to call itself Java.
Edited 2005-12-21 22:16
Sun offers a TCK ‘scholarship’ to qualified individuals and organisations. I’m in the process of working out the details with some very nice people at Sun regarding a TCK scholarship license for J2SE 1.5 for the GNU/*/Kaffe stack (Kaffe VM + GNU Classpath + various other missing bits and pieces to full 1.5 support), though I must admit that I have spent the last few months preparing the ground for the code to start testing [1], rather than nailing down the specifics of the legal agreements. First the code needs to get testable, then we can talk about modalities
I hope that we’ll have an agreement by mid next year, and can actually start certing GNU/Kaffe for compatibility. Obviously, such work would work best in concert with Harmony, and we’ll see how to make sure that collaboration on certification can happen. There is a lot of legalese involved, and trust building with Sun Microsystems, etc. We’ll see how it goes, I am pretty optimistic.
That being said, I’d hope that whichever free runtime turns out to be certed as compatible open the floodgates for a variety of other free runtimes to step into the same footsteps, as part of the strength of GNU Classpath is in the diversity of solutions, allowing you to hop from one GNU Classpath runtime to the next, in case it does not meet your goals. I see that a lot in embedded system scenarios where people evaluate different free runtimes for their workloads before settling on what work best in their scenario.
cheers,
dalibor topic
[1] And there are still a few things left to do until Kaffe is a great Classpath hacking environment, if you are interested in helping out, feel free to contact me.
“That being said, I’d hope that whichever free runtime turns out to be certed as compatible open the floodgates for a variety of other free runtimes to step into the same footsteps”
As long as it is 100% compatible, I am fine with that. But it has to be 100% compatible. That is the huge issue here. The last thing I want as a Java developer is people complaining to me that my applications have bugs, only to find out that it is not a bug in my app at all, but it is because they are running it on some JVM that is not 100% compatible.
I like the idea of an open source JVM. But I question the need for it. And I think that any open source JVM that wants to be truely accepted will have to be certified by Sun as 100% compatible. If it isn’t certified as 100% compatible, most users (including me) will not use it, and will keep using the Sun JVM instead. And I suspect, even if it is certified, the vast majority of users will keep using the Sun JVM unless there is some very compelling reason to switch.
“The last thing I want as a Java developer is people complaining to me that my applications have bugs, only to find out that it is not a bug in my app at all, but it is because they are running it on some JVM that is not 100% compatible.”
A lot of projects explicitely recommend specific VMs and vendors with which they have tested their work. I guess that is in part due to little differences in VMs (say, the 1.4 memory model’s inadequate specification leaving interesting opportunities for divergence) that can lead to different side effects at run time. That’s not a free VM specific feature, though, I ocassionally come across Java code that works on Sun’s VM but doesn’t on IBM’s etc. Write once, test everywhere, if you want to make sure that you know what your users are seing. It is usually not as bad as that sounds, though, using bytecode saves a recompilation step.
“And I suspect, even if it is certified, the vast majority of users will keep using the Sun JVM unless there is some very compelling reason to switch.”
Probably. Most people don’t use Linux or OpenSolaris (to name a Sun operating system project) because some other, proprietary platform is good enough for them. Some people find something that makes switching over worthwhile, others don’t. If you are not in it for the money, whether people switch or not is not very relevant. And I doubt there is a lot of money in gratis VMs.
It is very unlikely that people for whom the current modus operandi of proprietary runtimes is good enough will switch over to something else that’s similar to what they have as long as it is not significantly better. But that does not matter for me, for example, as with the free runtime like gcj being increasingly ‘good enough’ for my needs (open office2, eclipse, …) , I don’t have to fall into the ‘proprietary Java trap’: I could still chose to use proprietary implementations if that’s what I really wanted[1], but for those of us who prefer to use free software, there will be less and less explicit need to deal with Sun’s (or IBM’s, or Bea’s, or …) proprietary software, just because some cool library or application happens to be written in the Java programming language.
That’s where the major strength lies for the free implementations gcj: they are liberating code written in the Java programming language from the implicit constraint that users and developers have to deal with some proprietary piece of software to get it to build and work, which is a good thing, as it makes such code more useful in various scenarios where the proprietary software is an unadequate solution.
cheers,
dalibor topic
[1] Otoh, ruby is open source (GPL), has multiple implementations, and the recent crop of Java-to-ruby switchers are cheering on an incompatible ruby implementation called JRuby, rather than hacking on the real thing, as the incompatible implementation allows them to more easily interface with their legacy java code. And those switchers are from the good old J2EE and ‘pure-Java’ school of thought, going through funny epiphanies while trying to make their world views fit within the new reality they find themselves in.
People tend to be pragmatic in the majority: where free software offers cold hard cash worthy benefits over non-free software, like being able to sling out web sites 10 times faster using a fully free software stack, they tend to try to do that. Free markets can be odd that way.
Those who hope Sun will open-source all of its software products anytime soon are in for a big disappointment.
So they’re an open source company, they’re just not releasing their source code at this time. But anytime soon and they’ll be a big open source friendly business. You just wait.
Did I miss something, or didn’t Sun already realease a whole bunch of code?
You know, unimportant little things, like Open Office, or Solaris…
Yep, you missed something. Like this whole FOSS thing. Lots of source code released long before Sun became so generous.
But Sun ain’t giving source code because they believe in FOSS or because they’re trying to be nice. They’re giving source code to maintain their competitive image.
If it weren’t for them adopting AMD’s Opterons their system’s price/performance ratios would totally suck. They’re barely making a profit, if you can call it that, by still trying to be the ol’ proprietary Sun. The Dot in DotCom. Its obvious to anyone but Schwartz why those methods fail.
If Sun went closed source and UltraSparc all the way they would disappear from the market. They have no choice but to release source code and sell PCs.
“But Sun ain’t giving source code because they believe in FOSS or because they’re trying to be nice. They’re giving source code to maintain their competitive image.”
And you think IBM is any different? The only difference is that IBM recognized it would have to play nice with the FOSS community before Sun did. But now that Sun has recognized it, they are going quite a bit further than IBM when it comes to open sourcing their stuff.
But if you think any of the commercial companies adopting FOSS are doing it cause they are trying to be nice, you are kidding yourself. All of them are doing it only to gain a competitive advantage.
And you think IBM is any different?
Did I mention IBM?
No, you don’t think IBM is different. I can accept that. I haven’t seen much proof either way, except IBM did put up a huge legal defense to help Linux vs. SCO. IBM also gave a lot of enhancements to the Linux kernel, gcc, etc. And they promoted Linux via paid advertisements, some of them even of questionable legality. All in good fun.
So I know IBM is different than Sun in at least one way. They encourage Linux.
If Sun has a bad image among the Linux community it is probably deserved.
“except IBM did put up a huge legal defense to help Linux vs. SCO.”
Well, considering IBM was the one named in the lawsuit, and the one who was notified by SCO that their Unix license was terminated and they could no longer sell AIX, IBM didn’t exactly have a choice except to put up a legal defense. They only did it protect themselves. Not becuse of any good nature towards Linux.
“So I know IBM is different than Sun in at least one way. They encourage Linux.”
IBM is very good about speaking out of both sides of their mouth. They publically say they encourage Linux, but try to talk their customers into going AIX instead of Linux.
At least Sun is honest about it. They are honest about pushing their own Solaris operating system. But, Sun will preload Red Hat or SuSE on their AMD servers instead of Solaris if you ask them to.
And actually, Sun is the largest Linux vendor in the world because of their enormous deal Linux based Java Desktop System deal with China, which is on the order of 200 million copies.
“If Sun has a bad image among the Linux community it is probably deserved.”
Not really. It’s just that the typical FOSS zealot is a sheep who believes whatever they are told by prominant demigods in the Linux commnity like ESR and RMS. If ESR or RMS says Sun is bad, and IBM is good, than that is what the typical FOSS zealot will parrot back, along with the parroting back the rhetoric from ESR or RMS about why it is so. Most of the FOSS zealots don’t think for themselves, or create their own informed opinions. They just follow what people like ESR and RMS tell them.
The reality is that the vast majority of the Sun hatred in the FOSS community is because of one simple issue. Sun won’t give them Java under the GPL or LGPL. That’s really what it mostly comes down to: The militants in the FOSS community saying “We want Java as GPL or LGPL, and if you don’t give it to us, we are going to demonize you.”
Unfortunately, there are a lot of people in the FOSS community who think that this is the way to get companies to open source stuff they want. By threatening them and demonizing them if they don’t give them what they want.
Edited 2005-12-22 07:31
“Unfortunately, there are a lot of people in the FOSS community who think that this is the way to get companies to open source stuff they want. By threatening them and demonizing them if they don’t give them what they want. ”
You left out “duplicate and distribute…free”.
I’d rather they kept their code locked up tight in a little gold safe in the middle of their huge money bin. Makes it easier to forget about that way.
What incentive do they have to give technology to anyone? That big bin of money is the only reason they have for doing anything.
I had tried Gcj, not sure whether it works with eclipse or other java apps, but it pretty hard to use now.
No? see how they had compiled ant or eclipse.
It is distro-oriented, not source code-oriented..
Sometimes choice is no-good, but dogfood.
Simba, check out these screenshots:
http://gnu.wildebeest.org/diary/index.php?p=123
http://lists.gnu.org/archive/html/classpath/2005-10/msg00071.html
And those screenshots are pretty old now.
There is Swing support in the works; and those “90%” are not just about the core API.
The problem is that AFAIK, the FreeSwing is based on an obsolete Swing API.
That’s not correct. GNU Classpath has the goal to implement all of 1.4/1.5 APIs correctly, so it is not being based on some obsolete Swing API [1], but instead impleemnting the real thing. Not all components are there and finished yet, but it certainly is not impementing outdated APIs. There would be little point in limiting onself to J2SE 1.2 or somethig like that, when modern free software applications use 1.3/1.4.
cheers,
dalibor topic
[1] I can only guess what you mean here, unless you are alluding to the previous package name of javax.swing?
As the person cited I have to point out that the spin placed on my comments by the journalist is incorrect. Sun will be open sourcing everything it can just as fast as possible. I’ve written about his on my blog.
http://blogs.sun.com/roller/page/webmink?entry=fast_but_best_too
No rush…….to do anything.