They have no reason to make an open sourced version because they are perfectly happy with the set of open sourced tools they have now.
I think this is a shame, as I believe many don't know what they are missing. I am a fan of Java, but I would give it up quickly if there was any OS equivalent. My requirements are, I believe, simple. I want an OOP language with real performance. I don't want to have to mess about with dangerous memory management as with C or C++. I want binary compatibility on different platforms. I want good IDEs and tools. I want to be able to switch between different sources and vendors for my products. Is supporting open source a reason to put up with anything less? Not in my opinion.
That last group is still plugging along though, more power to them.
Absolutely. From what I hear, the open source Java developers are making good progress, with Java 1.5 and Swing support likely to appear in the near future.
"The GUI libraries can make use of DirectX and OpenGL."
They can or they do? On Windows Swing has an actual Win32 Window at the bottom and layers other graphic objects on top written in Java. Where does DirectX come into the picture?
Because DirectX can be used within Win32 Windows.
"The threading is superb"
Which threading? The Java threading behavior isn't the same on all platforms and isn't even consistent from one release of Java to the next.
These days Java threading is pretty much the same on all major platforms. It used to be very different (I remember the poor situation with early Linux versions). The threading is not consistent from release to the next because things improve. I would certainly not want the early bugs in threading continued on to later releases. The threading and concurrency facilities in modern Java is very good. Check out http://java.sun.com/j2se/1.5.0/docs/guide/concurre ncy/
When you use threading in Java you have to write your code as if your thread could be preempted at any time but also write it as if your thread will never be preempted.
Explain. Java has clear guidelines about what can be pre-empted and what can't (this is one of the reasons why porting Java onto 16-bit OSes can be hard, as all 32-bit operations are supposed to be atomic).
Sun FAILED to make java a viable desktop app platform. They also failed to make java a Free (in a GNU way) platform, and in doing this alienated a whole lot of would be users/developers. Also, this made java less platform independent than it could be if it was open-source, and less advanced.
Sorry, but this is just a poor excuse. Java is not a product; it is a specification. There is nothing to prevent clean-room open source Java versions from being developed and passing the compatibility tests.
If supporters of open source Java want a Free platform, thre is nothing to stop them writing one! In fact, there are many projects to do just that - such as GNU Classpath and Harmony.
So if you want to complain, complain about the rate of progress of those projects; don't blame Sun for not releasing their source code.
It's not realistic to believe that such a language could compete on the desktop with languages/libraries that are optomized for the hardware they run on.
Eh? Java's libraries are optimized for the hardware they run on. The GUI libraries can make use of DirectX and OpenGL. The threading is superb. All Java bytecode is run-time optimised to make good use of the underlying processor architecture.
Do you know why people are writing AJAX applications? It's because Sun failed the promise of java.
No, because Microsoft sabotaged having a quality JVM bundled on the client.
Do you know why there are so few java gui apps? Because Sun failed java.
There aren't few Java GUI apps. This is a common myth. Swing is used by a very large number of developers for internal client-side GUI apps within organisations. One of the most rapidly growing areas of Java is Rich Client development using the built-in resources of IDEs such as Eclipse and NetBeans.
Do you know why Ruby on Rails exist? Because Sun failed Java.
No, it is because many developers prefer open source development with languages like Ruby. Good for them. It a nice language. However, in terms of commercial impact, Ruby on Rails barely exists.
Java was all full of promise. Cross platform, run from the browser, free yourself from the drudgery of writing stateless apps using http and and that abortion known as javascript, no more learning 15 different gui toolkits, etc.
Well, that works for me. I don't use Javascript. I use Java-based tools like JSF to render it transparently for me.
So yes there are a billion java programmers all writing web apps but it's become a ghetto. Java was destined for bigger things.
Like what? Client side development? It is there. Numerical work? There. Real-time and embedded use? There. Mobile devices? There.
Lisp with declarations was just as fast as Fortran and C.
Yes - there are fast LISP implementations. But, yet again, what Java provides is multi-vendor And like Lisp with declarations, Java fails to solve the hard problems.
Like what?
You can write fast loops in it, but speed alone is insufficient for high-performance numerical work--Java is simply unsuitable for serious numerical work.
Factually incorrect. Java is being used for serious numerical work right now. For example, it is one of the supported languages on the University of Edinborough's supercomputer. Years ago, IBM produced research that showed that one of the Java implementations ran at FORTRAN-equalent speed.
That's typical: when people talk about implementations of the Java platform, people like you try to quietly change the subject and slip in "JVM". When people explicitly talk about J2SE, you try to confuse the issue by making remarks about J2ME. And TowerJ is an example of how successfully Sun has eliminated competing implementations--I suggest you look up what happened to it (TowerJ also never was a clean-room implementation of the Java platform--they relied on Sun's libraries).
Ah - goalpost moving! Change the language to mean the 'platform'. If you look at IBM's implementation, they use IBM code in place of sun.com. and sun.* libraries. So much so that when Sun wrote bad code in some products that used their JRE libraries, they broke with the IBM JRE!
The fact remains: there is no certified clean-room implementation of the J2SE platform; all conforming J2SE implementations are licensed derivatives of Sun's code.
The fact remains that there is and gave been. HP produced clean room full Java platform implementations.
Actually, it's you who is factually incorrect on many things,
Sounds a bit like the Emperor in Return of the Jedi:)
and who is using the distortions and misrepresentations typical of Java zealots. What you keep saying is what Sun promised (good numerical support, high performance generics, multiple independent implementations, good desktop implementation, universal in-browser delivery, open standards,...), but Sun has failed to deliver. I don't know whether people like you simply have trouble keeping reality and marketing apart, or whether you are deliberately astroturfing.
Look, why not quote facts? Paragraphs like this are meaningless. I have provided direct proof of errors in your statements (such as 'Java isn't suitable for numerical work'). Why not come back with actual facts?
In any case, I used to worry about this, but given that it looks like Java is going to remain confined to its niches, that Sun is going nowhere fast as a company, and that better alternatives to Java are widely available now, it simply doesn't matter much anymore.
Why not take a look at things like the TIOBE index, or the number of Java projects on sourceforce? Java is certainly not a niche language, and to claim it is blatantl nonsense, and it does not depend on Sun. If Sun vanished tomorrow, there are major companies with a huge investment in Java that will ensure it's long-term survival - IBM, BEA etc.
Sorry, but nyone who things that Java doesn't matter must be sadly out of touch with current IT and development. You may dislike it - I can understand that, but to claim it 'doesn't matter' is to be simply blinkered.
You are right in that reasonable alternatives may appear - particularly if Mono shows some performance improvements. However, in terms of support and tools, Java is currently dominant. In terms of high-performance server-side development, Java and J2EE are the key technology, and (for better or worse - I admit there are bad aspects) are likely to remain so for some time.
Algol 68 certainly had garbage collection, had a type system that was reasonably safe. It also had things that were roughly like things you would also call "classes" and "objects" in C++, but (like their equivalents in C++) no dynamic method binding.
Not garbage collection in the modern accepted sense. You had to manually release resources. It was certainly not an OOP language in any currently accepted way.
In any case, the point is that none of the individual features of Java were new, and you could get most of them in combination in prior languages. Other examples of very Java-like languages are Eiffel, ObjectPascal, Turing, Modula-3, and Simula 67.
This is true, but 'most of them' is not the same as 'all of them' - it is a crucial difference.
Java performance is no better than other safe languages with declared primitive types and arrays of primitive types in the past (actually, Java is a little worse because the garbage collector still sucks in the only official implementation). And like most other such languages, Java fails to address the hard cases: efficiency when using complicated abstractions, and genericity involving primitive types and user defined classes. It's the same stupid mistake that CommonLisp, Smalltalk, and other languages made. In fact, the only widely used platform that gets this right is C#/.NET 2.0.
This is factually incorrect. Java performance is substantially better than those equivalent languages - it approaches or matches C speed, which is something that was almost never the case for such other languages, at least, in their cross-platform implementations.
I know this because I was waiting for such a performance for years - I was not prepared to sacrifice C-equivalent speed to get portability and sensible memory management. I occasionally do high-performance numerical work. Java is the first 'safe' OOP language with garbage collection that I know will give me good cross-platform performance.
Yes, and that's a problem, because it has been Sun's strategy for killing competitors. The end result? A situation like Windows, in which we have a single vendor shipping and sublicensing a single, ill-defined platform that nobody else can implement fully.
Again, factually incorrect. Many others have fully implemented the JVM. TowerJ for example is a certified Java that does not rely on Sun's code. HP have produced clean-room certified Java.
Yes, and that kind of nonsense is exactly why Java has failed on the desktop: people want packages for their platform, they don't want a JAR file that doesn't integrate with the desktop.
Just because you can ship in a binary portable platform-independent way, doesn't mean you are forced to. There are, of course, Java installation packages for specific platforms. Just have a look at how InstallShield uses Java.
A GUI of poor quality and with poor desktop integration.
Sorry, but it really isn't poor quality. These days it is Direct/X and OpenGL accelerated, and has very good desktop integration (with JDIC). On future versions of Windows, for example, Swing applications will match the look and feel of the GUI to the exact pixel.
Useless because the market demands standard installers, so you have to repackage anyway. Also, other languages have provided this in the past.
Firstly, if the market (I assume you mean the desktop) demands standard installers, you can use them. Secondly, what market? The merest fraction of software development is for external shrink-wrapped use. Almost all of it is for internal use, where the ability to deploy a JAR (and update by simply replacing a JAR) is a phenomenal benefit. The majority of Java applications these days are web-based. And, there is indeed a standard way of installing those which is accepted by all vendors. It is the WAR file; a JAR with some additional meta-data in. Thirdly, as I keep saying, it matters not what other languages have provided in the past. The fact
So failing to find a flaw in his point, you just point out that one of his listed examples was out of place? You can conveniently ignore the very ancient LISP if you like, but it doesn't help your argument. His point was that the oh-so-special memory management features have existed for a long time, so your "for example" statement is just a red herring to distract away from the fact that you've been proven wrong.
Nowhere have I been proven wrong. My point was that Java combined the strengths of these languages with performance.
You completely forgot installation of the Java runtime in the first place. It's quite a pain to set up,
Yeah - single click downloads. Big deal.
especially on platforms like the BSDs which Sun doesn't care about as much.
Well, BSD are going to ship Java.
And don't even get me started on the mediocre "open" source alternatives.
I agree with you there - but they aren't Java - that is the point of certification.
Oh yeah and garbage collection has been available for C for a long time.
Yes, but that does still now allow binary portability, or remove the problems of pointer errors, does it?
After that argument failed, you disingenuously "moved the goal posts," so to speak. Your original argument was about these supposedly great and unique memory management features, now you're adding on other constraints retroactively. Learn to be a good sport and admit when your argument is invalid.
Please point out a single way in which my argument was invalid. I never said that these were unique memory management features. Of course they aren't. My point was that Java has provided them for mass use. Try and point to millions of developers using Lisp or Smalltalk.
Furthermore, you keep on towing that old line "but java is just as fast...really...trust me...I say so! Look, I have a few microbenchmarks (ignore the others) that showing it perform almost as good!"
No, actually. It is in the microbenchmarks that Java does not show performance. It is generally in substantial programs that run for reasonable lengths of time that Java picks up speed, due to the heavy run-time optimisation.
I hope you're just lying, because your criteria has been met for decades. Idiot.
No, they really haven't. Please name a single alternative language that was OOP, garbage collected, binary cross-platform, and high performance and was available from multiple vendors.
Translation: you write horrible code. Endian issues are trivially easy to deal with, as any network programmer will tell you. THere are standard functions for dealing with exactly that, how convenient! Java has them too, because it's necessary to communicate with standard internet protocols.
I am talking about the Joe Sixpack developer who simply wants to save data. You know - like open a file and write out data? Why on Earth should they have to use libraries? Java doesn't need such libraries.
The mere fact that there are libraries to deal with this indicates that there is an issue!
What is a "mutually text format"? Speak English please.
I can usually tolerate stupid rudness, but this is going a bit far. You criticise me for pointing out a major technical error in a previous post ('Algol is OOP') and then correct me for grammar. That is plain offensive.
If you do that, then you're stuck dependant on Java. Aside from being inherently less effecient (having the OS serialize entire objects for you), it also destroys the chance of an open standard.
When Java serialises objects, it is not less efficient. The OS is not serialising entire objects - Java libraries are.
Truly binary formats are easy, they're made all the time. In fact, they can and have been the standard for a long time. Pick your type of file and I'll point to how the vast majority are binary formats.
Given that Sun is a business, "succeeding" would mean earning a healthy profit on all the money invested in creating and promoting Java.
That was not the context of the orginal post, which suggested that they had failed Java in some unspecified way. You are discussing whether or not Java had failed them, which is a completely different matter.
But anyway, for all we know, they are earning a healthy profit from Java (their revenue has increased dramatically since the same quarter last year). What matters to them is where the losses are coming from. This may, or may not, be partly due to Java.
Now that their OS business is a lost cause sun should release the java sources under a license which lets people port it to different platforms. The user base will increase and they may be able to compete with C#
The user base is already huge, and it is competing against C# extremely well right now.
Despite all that, the company has really screwed up. I don't think they did a good job advocating Java or buying the mindshare of the development community.
Eh? Have you any idea of the size of the Java development community?
Java from SUN, who have failed Java like only SUN could.
I don't know what planet you are on, but on mine Java is one of the most successful and widely used development languages of all time. If that is 'failing', I would be interested in your definition of 'succeeding'.
What's your point? Algol, Smalltalk, and Lisp were safe, garbage-collected, object-oriented languages in widespread use before either C or C++ ever became popular, let alone Java. Java's contribution to the world of programming languages is exactly nothing.
Sorry, but this is wrong. For example Algol was neither safe, garbage-collected or object oriented.
Java contribution is huge, in that it took what these other (excellent) languages offered and finally packed it together into a single system that was practical.
I know this, because I have developed in a large number of languages (including Algol (60 and 68), Smalltalk and LISP). None of these languages combined high performance, multi-vendor support for the same dialect, good performance and cross-platform portability at the binary level (and with a cross-platform gui) and also being free (as in beer).
I had been waiting for a language like that for literally decades. Smalltalk came closest, but there was never a cheap cross-platform version that came within better than 20-30% of the performance of C.
Processor, word size, and OS dependencies, they are not an issue even in correctly written C/C++, and certainly not in most other languages.
This simply is false. As a past C/C++ developer, I had to deal with such issues all the time. Writing truly portable binary data formats with C or C++ (or other widely used languages such as Pascal or Fortran) was a major issue, which is why so many proprietary and mutually text formats for specialised use were developed. There is also the matter of running distributed code. Efficiency requires binary messaging; where word size and processor features really can screw things up.
In Java, if you serialise a binary object, you can re-load it on any platform, 32-bit, 64-bit, big endian, little endian, or whatever. This has phenomenal advantages.
And even in Java, developers still have to build and maintain binaries for different processor architectures, because properly packaging applications for different platforms requires that.
No, it doesn't. There are standard packaging methods such as Jar and WAR which are fully platform independent. I have deployed identical JARs on Windows and Linux, and MacOS/X. The only thing that matters is how you specify the classpath in different OSes, and that is the merest fraction of effort (a few lines of script). All you need is a single script for each platform which can be supplied with the name of the main Jar and the directory containing library jars, and that is all you ever need for any application for that platform.
Actually things can be even simpler than that! There are java utilities like 'one-jar' that allows all Jars for a project to be deployed and run as a single binary archive for any platform.
But while C++ sucks, at least people regularly write high quality desktop apps in it, which is more than one can say for Java.
You would be surprised. Azeurus is Java. Moneydance (one of the most popular personal accounts packages) is Java. Java is now even starting to be used for PC games! (Tribal Trouble!)
The desktop app situation is a specialised one, as developers usually haven't cared about portability, and OS vendors have sold toolkits for development in C or C++. However, now you can get a wide range of portable Java GUI APIs - Swing, SWT, Qt, GTK+. Of course, with all but Swing you lose the ability of clean cross-platform binary deployment.
1. Look at the number of people who are still using Microsoft Office 97. Why don't they upgrade when "better" versions are available? Because the version they have fulfills their needs.
2. Similarly, the number of companies still using Exchange 5.5 is staggeringly high. Why haven't they upgraded? Because the version they have fulfills their needs.
There will always be some customers that want/need faster/better, but rarely can you build a $13 billion (Oracle's revenues last year) company on them.
These aren't the same type of market at all. Take the example of Office 97 - people aren't typing orders of magnitudes faster than 10 years ago, so there is no major need for increased functionality in terms of word processing. However, consider something like banking. We have moved from mass use of cheques and transactions at local branches to real-time 24-hour on-line banking. The demands on servers and databases has increased by orders of magnitudes. There are plenty of other examples where IT customers need faster and better.
What distinguishes Java from other cross-platform solutions is that you only have to compile once, but that's a nearly useless feature, and one that comes at a huge price in terms of quality and performance.
Nonsense. Even if was not cross platform, Java would be an important language as it removes the horrors of C/C++ memory management, and no-one with any experience of modern Java runtimes would state that there is a performance price. Java runs everywhere from embedded systems, mobile devices, real-time systems up to the highest level of clustered mainframe systems, and the runtime code analysis and machine code transation produces C-equivalent performance in all but the smallest embedded systems (where the small size of Java byte codes often gives it a memory advantage).
The ability to compile once is a tremendous benefit. No longer do developers have to build and maintain binaries for different processor architectures, word sizes and operating systems. Developers like me deploy cross-platform binaries on a daily basis. (The often stated 'Write One Run Anywhere is a myth' is itself a myth).
This is such a powerful feature that it allows clustering of heterogenous systems, with live code deployment from a single binary - I can cluster J2EE systems on say, Linux, Windows server and Solaris and provide a single binary WAR file which is automatically deployed to all of them.
It is this sort of thing that makes C++ look, in this respect, primitive.
You're both partially correct. While the high-end does increase as you say, it typically does it not increase linearly. Thus, the number of users that something like MySQL or PostgreSQL can satisfy will grow more quickly than the those that need or even want high-end features. This will leave an ever dwindling number of users whose needs Oracle can uniquely satisfy.
What evidence do you have that it increases in a particular manner?
Sorry, but I just don't believe the 'ever dwindling' argument. It seems to me to come from people who generally don't understand the requirements of serious commercial software. I really like postgresql, and use it a lot, but there comes a time (when I want very high uptime and garanteed commercial support) that I switch my applications to Oracle, or some equivalent. The price of these databases is insignificant compared with the potential costs if the application failed.
Oracle's biggest product is a high end database. As the performance of its open source competitors improves, that means there is less and less room for Oracle. In other words, Oracle gets chased up-market. Eventually it ceases to exist.
Flawed logic. What happens is that the demands of the high-end continually increase, and the performance of both open source competitors and Oracle improve at the same time. Oracle retains its margin. The things that oracle's database can cope with are awesome. I have heard of uses which involve tables consisting of thousands of columns, not merely thousands of rows. Sorry, but I really don't see PostgreSQL or MySQL dealing with this sort of demand in the near future.
A silly statement. Java is being used to develop applications that can be deployed cross-platform right now by a vast number of developers. Nothing comes close.
Every application (and even lots of low-level system functionality) that is included with every current OS fits this description. Why is it no-one gets upset about modern OSes "bundling" shells, network stacks, widget libraries, GUIs, printer drivers, CPU schedulers, text editors, calculators, FTP clients, card games, etc, etc, etc ?
Because Microsoft does not generally do deals to prevent other companies bundling alternatives to these. Microsoft's OEM licensing terms prevented many things like this. Microsoft targetted specifically things that would allow them to control access to major new forms of media or information.
There is all the world of difference between a card game and an internet browser or multimedia player. There is no substantial market for the distribution of card games!!
However, there are no independent, objective metrics. A company cannot know whether or not it is a monopoly until a court tells it.
Of course they can. As I said, they have substantial legal departments that can advise them that they are moving into dangerous areas in terms of monopolistic behaviour. To simply wait for a court to decide is incredibly dumb - any good legal team can advise that problems may lie ahead.
How many other OSes can you name that *don't* come with "media players, etc" ?
That is, of course, not the point. Firstly, those companies are not monopolies; secondly many of them bundle several alternatives.
And if none of the competitor's products provided sufficient/equivalent functionality ?
Well hard luck to them.
How can a company in a monopoly position provide *better* (or even equivalent) functionality than their competitors ? They cannot, because by doing so they will be considered "abusive".
Sorry, but I think you need to re-read my posts.
A company in a monopoly position can of course provide better functionality than their competitors.
But just not a bundled products. There is nothing to stop Microsoft providing the best quality media player ever.
Or, as I said, a company in a "monopoly" can no longer compete, nor deliver new features - in line with competitor's products - to its customers, because it will be considered "abusive".
Delivering features in something that is an operating system is one thing. Bundling things that had generally been accepted as not being part of an OS before (such as a browser) is obviously and blatantly something else.
So why did (and does) the same logic not apply to CPU schedulers, network stacks, memory management, hardware drivers, widget sets, text editors, calculators, disk repair tools, compression tools, backup utilities, shells, GUIs, etc, etc.
Ideally, it should; and there have been exactly such debates over aspects of Windows in the past - matters such as Netware's stack, Diskeeper's NTFS disk tools etc.
There are different issues with such things, however, as what Windows normally does is provide low-functionality basic versions of many of these things (such as disk repair tools), which don't really compete with the specialised and high quality tools offered by competitors. This means that the competitors can still make substantial amounts of money.
However, what we are dealing with here is far, far larger markets - the distribution of a wide range of media types, to the extent that Windows is in some products being sold as an entire home media centre, for films, music, TV etc.
Surely you can see that there is a difference between bundling a notepad text editor and an entire home media centre. That is what Microsoft are trying to do: to leverage their substantial (monopoly) desktop presence to gain power in other markets.
That is illegal.
Microsoft have not "bundled" anything that their competitors weren't already, or subsequently haven't. Indeed, one of the standard criticisms of Windows on Slashdot is that it doesn't come with *enough* functionality.
When is the point goint to get through? This matters because they are a monopoly.
Microsoft have not "bundled" anything that their competitors weren't already, or subsequently haven't.
Nonsense. For example, back in the early 90s, Internet Explorer was not bundled. It was shipped on separate media. Microsoft deliberately tied it in as part of the core OS to crush Netscape. That is established fact.
Linux, but at least it's usually free). I have _zero_ interest in having to do that again. Most people think OSes gaining more and more functionality out of the box is a *good* thing.
Firstly, you can of course have more functionality out of the box. But why not do it the Linux way, and have more than one alternative provider of that functio
Slashdot Mirror
Quit complaining and just write in a real language.
Perl/Tk
Wow yes - I would love to try and write my image processing software, numerical application, or database in that.
They have no reason to make an open sourced version because they are perfectly happy with the set of open sourced tools they have now.
I think this is a shame, as I believe many don't know what they are missing. I am a fan of Java, but I would give it up quickly if there was any OS equivalent. My requirements are, I believe, simple. I want an OOP language with real performance. I don't want to have to mess about with dangerous memory management as with C or C++. I want binary compatibility on different platforms. I want good IDEs and tools. I want to be able to switch between different sources and vendors for my products. Is supporting open source a reason to put up with anything less? Not in my opinion.
That last group is still plugging along though, more power to them.
Absolutely. From what I hear, the open source Java developers are making good progress, with Java 1.5 and Swing support likely to appear in the near future.
"The GUI libraries can make use of DirectX and OpenGL."
e ncy/
They can or they do? On Windows Swing has an actual Win32 Window at the bottom and layers other graphic objects on top written in Java. Where does DirectX come into the picture?
Because DirectX can be used within Win32 Windows.
"The threading is superb"
Which threading? The Java threading behavior isn't the same on all platforms and isn't even consistent from one release of Java to the next.
These days Java threading is pretty much the same on all major platforms. It used to be very different (I remember the poor situation with early Linux versions). The threading is not consistent from release to the next because things improve. I would certainly not want the early bugs in threading continued on to later releases. The threading and concurrency facilities in modern Java is very good. Check out http://java.sun.com/j2se/1.5.0/docs/guide/concurr
When you use threading in Java you have to write your code as if your thread could be preempted at any time but also write it as if your thread will never be preempted.
Explain. Java has clear guidelines about what can be pre-empted and what can't (this is one of the reasons why porting Java onto 16-bit OSes can be hard, as all 32-bit operations are supposed to be atomic).
My planet is the one where C is still the most successful and widely used.
Well, it is not my planet. Check out the TIOBE index of software resources.
It is the one where Java failed utterly in the web client space, despite having a tremendous lead.
yes, thanks to sabotage by Microsoft.
It is the one where Java loses ground to PHP, python, and now Ruby, by the hour.
You must be joking! Have you actually checked the job lists for these skills? I suggest you do before you make such wild statements.
It is the one where java has failed on the desktop nearly completely.
A common myth. Java is widely used on the desktop. Just because you don't see it in many shrink-wrapped applications, does not mean it is not there.
Sun FAILED to make java a viable desktop app platform. They also failed to make java a Free (in a GNU way) platform, and in doing this alienated a whole lot of would be users/developers. Also, this made java less platform independent than it could be if it was open-source, and less advanced.
Sorry, but this is just a poor excuse. Java is not a product; it is a specification. There is nothing to prevent clean-room open source Java versions from being developed and passing the compatibility tests.
If supporters of open source Java want a Free platform, thre is nothing to stop them writing one! In fact, there are many projects to do just that - such as GNU Classpath and Harmony.
So if you want to complain, complain about the rate of progress of those projects; don't blame Sun for not releasing their source code.
It's not realistic to believe that such a language could compete on the desktop with languages/libraries that are optomized for the hardware they run on.
Eh? Java's libraries are optimized for the hardware they run on. The GUI libraries can make use of DirectX and OpenGL. The threading is superb. All Java bytecode is run-time optimised to make good use of the underlying processor architecture.
Do you know why people are writing AJAX applications? It's because Sun failed the promise of java.
No, because Microsoft sabotaged having a quality JVM bundled on the client.
Do you know why there are so few java gui apps? Because Sun failed java.
There aren't few Java GUI apps. This is a common myth. Swing is used by a very large number of developers for internal client-side GUI apps within organisations. One of the most rapidly growing areas of Java is Rich Client development using the built-in resources of IDEs such as Eclipse and NetBeans.
Do you know why Ruby on Rails exist? Because Sun failed Java.
No, it is because many developers prefer open source development with languages like Ruby. Good for them. It a nice language. However, in terms of commercial impact, Ruby on Rails barely exists.
Java was all full of promise. Cross platform, run from the browser, free yourself from the drudgery of writing stateless apps using http and and that abortion known as javascript, no more learning 15 different gui toolkits, etc.
Well, that works for me. I don't use Javascript. I use Java-based tools like JSF to render it transparently for me.
So yes there are a billion java programmers all writing web apps but it's become a ghetto. Java was destined for bigger things.
Like what? Client side development? It is there. Numerical work? There. Real-time and embedded use? There. Mobile devices? There.
Lisp with declarations was just as fast as Fortran and C.
:)
...), but Sun has failed to deliver. I don't know whether people like you simply have trouble keeping reality and marketing apart, or whether you are deliberately astroturfing.
Yes - there are fast LISP implementations. But, yet again, what Java provides is multi-vendor
And like Lisp with declarations, Java fails to solve the hard problems.
Like what?
You can write fast loops in it, but speed alone is insufficient for high-performance numerical work--Java is simply unsuitable for serious numerical work.
Factually incorrect. Java is being used for serious numerical work right now. For example, it is one of the supported languages on the University of Edinborough's supercomputer. Years ago, IBM produced research that showed that one of the Java implementations ran at FORTRAN-equalent speed.
That's typical: when people talk about implementations of the Java platform, people like you try to quietly change the subject and slip in "JVM". When people explicitly talk about J2SE, you try to confuse the issue by making remarks about J2ME. And TowerJ is an example of how successfully Sun has eliminated competing implementations--I suggest you look up what happened to it (TowerJ also never was a clean-room implementation of the Java platform--they relied on Sun's libraries).
Ah - goalpost moving! Change the language to mean the 'platform'. If you look at IBM's implementation, they use IBM code in place of sun.com. and sun.* libraries. So much so that when Sun wrote bad code in some products that used their JRE libraries, they broke with the IBM JRE!
The fact remains: there is no certified clean-room implementation of the J2SE platform; all conforming J2SE implementations are licensed derivatives of Sun's code.
The fact remains that there is and gave been. HP produced clean room full Java platform implementations.
Actually, it's you who is factually incorrect on many things,
Sounds a bit like the Emperor in Return of the Jedi
and who is using the distortions and misrepresentations typical of Java zealots. What you keep saying is what Sun promised (good numerical support, high performance generics, multiple independent implementations, good desktop implementation, universal in-browser delivery, open standards,
Look, why not quote facts? Paragraphs like this are meaningless. I have provided direct proof of errors in your statements (such as 'Java isn't suitable for numerical work'). Why not come back with actual facts?
In any case, I used to worry about this, but given that it looks like Java is going to remain confined to its niches, that Sun is going nowhere fast as a company, and that better alternatives to Java are widely available now, it simply doesn't matter much anymore.
Why not take a look at things like the TIOBE index, or the number of Java projects on sourceforce? Java is certainly not a niche language, and to claim it is blatantl nonsense, and it does not depend on Sun. If Sun vanished tomorrow, there are major companies with a huge investment in Java that will ensure it's long-term survival - IBM, BEA etc.
Sorry, but nyone who things that Java doesn't matter must be sadly out of touch with current IT and development. You may dislike it - I can understand that, but to claim it 'doesn't matter' is to be simply blinkered.
You are right in that reasonable alternatives may appear - particularly if Mono shows some performance improvements. However, in terms of support and tools, Java is currently dominant. In terms of high-performance server-side development, Java and J2EE are the key technology, and (for better or worse - I admit there are bad aspects) are likely to remain so for some time.
Algol 68 certainly had garbage collection, had a type system that was reasonably safe. It also had things that were roughly like things you would also call "classes" and "objects" in C++, but (like their equivalents in C++) no dynamic method binding.
Not garbage collection in the modern accepted sense. You had to manually release resources. It was certainly not an OOP language in any currently accepted way.
In any case, the point is that none of the individual features of Java were new, and you could get most of them in combination in prior languages. Other examples of very Java-like languages are Eiffel, ObjectPascal, Turing, Modula-3, and Simula 67.
This is true, but 'most of them' is not the same as 'all of them' - it is a crucial difference.
Java performance is no better than other safe languages with declared primitive types and arrays of primitive types in the past (actually, Java is a little worse because the garbage collector still sucks in the only official implementation). And like most other such languages, Java fails to address the hard cases: efficiency when using complicated abstractions, and genericity involving primitive types and user defined classes. It's the same stupid mistake that CommonLisp, Smalltalk, and other languages made. In fact, the only widely used platform that gets this right is C#/.NET 2.0.
This is factually incorrect. Java performance is substantially better than those equivalent languages - it approaches or matches C speed, which is something that was almost never the case for such other languages, at least, in their cross-platform implementations.
I know this because I was waiting for such a performance for years - I was not prepared to sacrifice C-equivalent speed to get portability and sensible memory management. I occasionally do high-performance numerical work. Java is the first 'safe' OOP language with garbage collection that I know will give me good cross-platform performance.
Yes, and that's a problem, because it has been Sun's strategy for killing competitors. The end result? A situation like Windows, in which we have a single vendor shipping and sublicensing a single, ill-defined platform that nobody else can implement fully.
Again, factually incorrect. Many others have fully implemented the JVM. TowerJ for example is a certified Java that does not rely on Sun's code. HP have produced clean-room certified Java.
Yes, and that kind of nonsense is exactly why Java has failed on the desktop: people want packages for their platform, they don't want a JAR file that doesn't integrate with the desktop.
Just because you can ship in a binary portable platform-independent way, doesn't mean you are forced to. There are, of course, Java installation packages for specific platforms. Just have a look at how InstallShield uses Java.
A GUI of poor quality and with poor desktop integration.
Sorry, but it really isn't poor quality. These days it is Direct/X and OpenGL accelerated, and has very good desktop integration (with JDIC). On future versions of Windows, for example, Swing applications will match the look and feel of the GUI to the exact pixel.
Useless because the market demands standard installers, so you have to repackage anyway. Also, other languages have provided this in the past.
Firstly, if the market (I assume you mean the desktop) demands standard installers, you can use them. Secondly, what market? The merest fraction of software development is for external shrink-wrapped use. Almost all of it is for internal use, where the ability to deploy a JAR (and update by simply replacing a JAR) is a phenomenal benefit. The majority of Java applications these days are web-based. And, there is indeed a standard way of installing those which is accepted by all vendors. It is the WAR file; a JAR with some additional meta-data in. Thirdly, as I keep saying, it matters not what other languages have provided in the past. The fact
So failing to find a flaw in his point, you just point out that one of his listed examples was out of place? You can conveniently ignore the very ancient LISP if you like, but it doesn't help your argument. His point was that the oh-so-special memory management features have existed for a long time, so your "for example" statement is just a red herring to distract away from the fact that you've been proven wrong.
Nowhere have I been proven wrong. My point was that Java combined the strengths of these languages with performance.
You completely forgot installation of the Java runtime in the first place. It's quite a pain to set up,
Yeah - single click downloads. Big deal.
especially on platforms like the BSDs which Sun doesn't care about as much.
Well, BSD are going to ship Java.
And don't even get me started on the mediocre "open" source alternatives.
I agree with you there - but they aren't Java - that is the point of certification.
Oh yeah and garbage collection has been available for C for a long time.
Yes, but that does still now allow binary portability, or remove the problems of pointer errors, does it?
After that argument failed, you disingenuously "moved the goal posts," so to speak. Your original argument was about these supposedly great and unique memory management features, now you're adding on other constraints retroactively. Learn to be a good sport and admit when your argument is invalid.
Please point out a single way in which my argument was invalid. I never said that these were unique memory management features. Of course they aren't. My point was that Java has provided them for mass use. Try and point to millions of developers using Lisp or Smalltalk.
Furthermore, you keep on towing that old line "but java is just as fast...really...trust me...I say so! Look, I have a few microbenchmarks (ignore the others) that showing it perform almost as good!"
No, actually. It is in the microbenchmarks that Java does not show performance. It is generally in substantial programs that run for reasonable lengths of time that Java picks up speed, due to the heavy run-time optimisation.
I hope you're just lying, because your criteria has been met for decades. Idiot.
No, they really haven't. Please name a single alternative language that was OOP, garbage collected, binary cross-platform, and high performance and was available from multiple vendors.
Translation: you write horrible code. Endian issues are trivially easy to deal with, as any network programmer will tell you. THere are standard functions for dealing with exactly that, how convenient! Java has them too, because it's necessary to communicate with standard internet protocols.
I am talking about the Joe Sixpack developer who simply wants to save data. You know - like open a file and write out data? Why on Earth should they have to use libraries? Java doesn't need such libraries.
The mere fact that there are libraries to deal with this indicates that there is an issue!
What is a "mutually text format"? Speak English please.
I can usually tolerate stupid rudness, but this is going a bit far. You criticise me for pointing out a major technical error in a previous post ('Algol is OOP') and then correct me for grammar. That is plain offensive.
If you do that, then you're stuck dependant on Java. Aside from being inherently less effecient (having the OS serialize entire objects for you), it also destroys the chance of an open standard.
When Java serialises objects, it is not less efficient. The OS is not serialising entire objects - Java libraries are.
Truly binary formats are easy, they're made all the time. In fact, they can and have been the standard for a long time. Pick your type of file and I'll point to how the vast majority are binary formats.
OK - chemical data structure files.
Wow,
Given that Sun is a business, "succeeding" would mean earning a healthy profit on all the money invested in creating and promoting Java.
That was not the context of the orginal post, which suggested that they had failed Java in some unspecified way. You are discussing whether or not Java had failed them, which is a completely different matter.
But anyway, for all we know, they are earning a healthy profit from Java (their revenue has increased dramatically since the same quarter last year). What matters to them is where the losses are coming from. This may, or may not, be partly due to Java.
No! That would kill Java.
Not necessarily. What matters is that distributions pass the compatibility tests. There is nothing to stop open source versions passing these tests.
Now that their OS business is a lost cause sun should release the java sources under a license which lets people port it to different platforms. The user base will increase and they may be able to compete with C#
The user base is already huge, and it is competing against C# extremely well right now.
Despite all that, the company has really screwed up. I don't think they did a good job advocating Java or buying the mindshare of the development community.
Eh? Have you any idea of the size of the Java development community?
Java from SUN, who have failed Java like only SUN could.
I don't know what planet you are on, but on mine Java is one of the most successful and widely used development languages of all time. If that is 'failing', I would be interested in your definition of 'succeeding'.
What's your point? Algol, Smalltalk, and Lisp were safe, garbage-collected, object-oriented languages in widespread use before either C or C++ ever became popular, let alone Java. Java's contribution to the world of programming languages is exactly nothing.
Sorry, but this is wrong. For example Algol was neither safe, garbage-collected or object oriented.
Java contribution is huge, in that it took what these other (excellent) languages offered and finally packed it together into a single system that was practical.
I know this, because I have developed in a large number of languages (including Algol (60 and 68), Smalltalk and LISP). None of these languages combined high performance, multi-vendor support for the same dialect, good performance and cross-platform portability at the binary level (and with a cross-platform gui) and also being free (as in beer).
I had been waiting for a language like that for literally decades. Smalltalk came closest, but there was never a cheap cross-platform version that came within better than 20-30% of the performance of C.
Processor, word size, and OS dependencies, they are not an issue even in correctly written C/C++, and certainly not in most other languages.
This simply is false. As a past C/C++ developer, I had to deal with such issues all the time. Writing truly portable binary data formats with C or C++ (or other widely used languages such as Pascal or Fortran) was a major issue, which is why so many proprietary and mutually text formats for specialised use were developed. There is also the matter of running distributed code. Efficiency requires binary messaging; where word size and processor features really can screw things up.
In Java, if you serialise a binary object, you can re-load it on any platform, 32-bit, 64-bit, big endian, little endian, or whatever. This has phenomenal advantages.
And even in Java, developers still have to build and maintain binaries for different processor architectures, because properly packaging applications for different platforms requires that.
No, it doesn't. There are standard packaging methods such as Jar and WAR which are fully platform independent. I have deployed identical JARs on Windows and Linux, and MacOS/X. The only thing that matters is how you specify the classpath in different OSes, and that is the merest fraction of effort (a few lines of script). All you need is a single script for each platform which can be supplied with the name of the main Jar and the directory containing library jars, and that is all you ever need for any application for that platform.
Actually things can be even simpler than that! There are java utilities like 'one-jar' that allows all Jars for a project to be deployed and run as a single binary archive for any platform.
But while C++ sucks, at least people regularly write high quality desktop apps in it, which is more than one can say for Java.
You would be surprised. Azeurus is Java. Moneydance (one of the most popular personal accounts packages) is Java. Java is now even starting to be used for PC games! (Tribal Trouble!)
The desktop app situation is a specialised one, as developers usually haven't cared about portability, and OS vendors have sold toolkits for development in C or C++. However, now you can get a wide range of portable Java GUI APIs - Swing, SWT, Qt, GTK+. Of course, with all but Swing you lose the ability of clean cross-platform binary deployment.
A couple of examples:
1. Look at the number of people who are still using Microsoft Office 97. Why don't they upgrade when "better" versions are available? Because the version they have fulfills their needs.
2. Similarly, the number of companies still using Exchange 5.5 is staggeringly high. Why haven't they upgraded? Because the version they have fulfills their needs.
There will always be some customers that want/need faster/better, but rarely can you build a $13 billion (Oracle's revenues last year) company on them.
These aren't the same type of market at all. Take the example of Office 97 - people aren't typing orders of magnitudes faster than 10 years ago, so there is no major need for increased functionality in terms of word processing. However, consider something like banking. We have moved from mass use of cheques and transactions at local branches to real-time 24-hour on-line banking. The demands on servers and databases has increased by orders of magnitudes. There are plenty of other examples where IT customers need faster and better.
What distinguishes Java from other cross-platform solutions is that you only have to compile once, but that's a nearly useless feature, and one that comes at a huge price in terms of quality and performance.
Nonsense. Even if was not cross platform, Java would be an important language as it removes the horrors of C/C++ memory management, and no-one with any experience of modern Java runtimes would state that there is a performance price. Java runs everywhere from embedded systems, mobile devices, real-time systems up to the highest level of clustered mainframe systems, and the runtime code analysis and machine code transation produces C-equivalent performance in all but the smallest embedded systems (where the small size of Java byte codes often gives it a memory advantage).
The ability to compile once is a tremendous benefit. No longer do developers have to build and maintain binaries for different processor architectures, word sizes and operating systems. Developers like me deploy cross-platform binaries on a daily basis. (The often stated 'Write One Run Anywhere is a myth' is itself a myth).
This is such a powerful feature that it allows clustering of heterogenous systems, with live code deployment from a single binary - I can cluster J2EE systems on say, Linux, Windows server and Solaris and provide a single binary WAR file which is automatically deployed to all of them.
It is this sort of thing that makes C++ look, in this respect, primitive.
You're both partially correct. While the high-end does increase as you say, it typically does it not increase linearly. Thus, the number of users that something like MySQL or PostgreSQL can satisfy will grow more quickly than the those that need or even want high-end features. This will leave an ever dwindling number of users whose needs Oracle can uniquely satisfy.
What evidence do you have that it increases in a particular manner?
Sorry, but I just don't believe the 'ever dwindling' argument. It seems to me to come from people who generally don't understand the requirements of serious commercial software. I really like postgresql, and use it a lot, but there comes a time (when I want very high uptime and garanteed commercial support) that I switch my applications to Oracle, or some equivalent. The price of these databases is insignificant compared with the potential costs if the application failed.
Oracle's biggest product is a high end database. As the performance of its open source competitors improves, that means there is less and less room for Oracle. In other words, Oracle gets chased up-market. Eventually it ceases to exist.
Flawed logic. What happens is that the demands of the high-end continually increase, and the performance of both open source competitors and Oracle improve at the same time. Oracle retains its margin. The things that oracle's database can cope with are awesome. I have heard of uses which involve tables consisting of thousands of columns, not merely thousands of rows. Sorry, but I really don't see PostgreSQL or MySQL dealing with this sort of demand in the near future.
A silly statement. Java is being used to develop applications that can be deployed cross-platform right now by a vast number of developers. Nothing comes close.
Why can't other constants, like say pi, be variable as well?
Pi is not a physical constant. It is the result of a mathematical expression. It can't change.
Why trust the EU to get this right and then not go after SuSE and RedHat for bundling only one or two players and integrating them into KDE and GNOME?
Because KDE and GNOME are not in a monopoly situation.
Every application (and even lots of low-level system functionality) that is included with every current OS fits this description. Why is it no-one gets upset about modern OSes "bundling" shells, network stacks, widget libraries, GUIs, printer drivers, CPU schedulers, text editors, calculators, FTP clients, card games, etc, etc, etc ?
Because Microsoft does not generally do deals to prevent other companies bundling alternatives to these. Microsoft's OEM licensing terms prevented many things like this. Microsoft targetted specifically things that would allow them to control access to major new forms of media or information.
There is all the world of difference between a card game and an internet browser or multimedia player. There is no substantial market for the distribution of card games!!
However, there are no independent, objective metrics. A company cannot know whether or not it is a monopoly until a court tells it.
.
Of course they can. As I said, they have substantial legal departments that can advise them that they are moving into dangerous areas in terms of monopolistic behaviour. To simply wait for a court to decide is incredibly dumb - any good legal team can advise that problems may lie ahead.
How many other OSes can you name that *don't* come with "media players, etc" ?
That is, of course, not the point. Firstly, those companies are not monopolies; secondly many of them bundle several alternatives.
And if none of the competitor's products provided sufficient/equivalent functionality ?
Well hard luck to them.
How can a company in a monopoly position provide *better* (or even equivalent) functionality than their competitors ? They cannot, because by doing so they will be considered "abusive".
Sorry, but I think you need to re-read my posts.
A company in a monopoly position can of course provide better functionality than their competitors.
But just not a bundled products. There is nothing to stop Microsoft providing the best quality media player ever.
Or, as I said, a company in a "monopoly" can no longer compete, nor deliver new features - in line with competitor's products - to its customers, because it will be considered "abusive".
Delivering features in something that is an operating system is one thing. Bundling things that had generally been accepted as not being part of an OS before (such as a browser) is obviously and blatantly something else.
So why did (and does) the same logic not apply to CPU schedulers, network stacks, memory management, hardware drivers, widget sets, text editors, calculators, disk repair tools, compression tools, backup utilities, shells, GUIs, etc, etc.
Ideally, it should; and there have been exactly such debates over aspects of Windows in the past - matters such as Netware's stack, Diskeeper's NTFS disk tools etc.
There are different issues with such things, however, as what Windows normally does is provide low-functionality basic versions of many of these things (such as disk repair tools), which don't really compete with the specialised and high quality tools offered by competitors. This means that the competitors can still make substantial amounts of money.
However, what we are dealing with here is far, far larger markets - the distribution of a wide range of media types, to the extent that Windows is in some products being sold as an entire home media centre, for films, music, TV etc.
Surely you can see that there is a difference between bundling a notepad text editor and an entire home media centre. That is what Microsoft are trying to do: to leverage their substantial (monopoly) desktop presence to gain power in other markets.
That is illegal.
Microsoft have not "bundled" anything that their competitors weren't already, or subsequently haven't. Indeed, one of the standard criticisms of Windows on Slashdot is that it doesn't come with *enough* functionality.
When is the point goint to get through? This matters because they are a monopoly
Microsoft have not "bundled" anything that their competitors weren't already, or subsequently haven't.
Nonsense. For example, back in the early 90s, Internet Explorer was not bundled. It was shipped on separate media. Microsoft deliberately tied it in as part of the core OS to crush Netscape. That is established fact.
Linux, but at least it's usually free). I have _zero_ interest in having to do that again. Most people think OSes gaining more and more functionality out of the box is a *good* thing.
Firstly, you can of course have more functionality out of the box. But why not do it the Linux way, and have more than one alternative provider of that functio