Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Java Programming

IBM releases JDK 1.16 alpha for Linux 66

Posted by sengan from the should-kick-ass dept.
Jeff N Roberts from IBM wrote in to announce that IBM is releasing a JDK for Linux which incorporates the same IBM Just-In-Time technology found in IBM's JVMs for OS/2 and Windows. Dave Whitinger has HTMLized the FAQ we were sent. (If the IBM link does not work, it should be active shortly)
This discussion has been archived. No new comments can be posted.

IBM releases JDK 1.16 alpha for Linux More

IBM releases JDK 1.16 alpha for Linux

Comments Filter:
  • Sure, use C++ and OpenGL for games, use
    Java when you really find a good use for it
    (as in servlets). Use what's best given
    the circumstances. Java is piss easy and
    has a brilliant API, but that doesn't make it
    as fast as C.

  • This discussion is pissing me off! (Score:1)

    by Anonymous Coward
    Guys and gals,

    Java is good, C is good (C++/MFC, naaa..;0)

    writing system software in C is the way it is done.
    writing web software in java is the way it is done.

    I am a professional and what I see i this: java on the server for web deployment is almost the de facto standard. We use servlets (apache) and EJBoss for session management. Web development is different than system development.

    Programming for the web is a different kind of programming than programming for systems. SO STOP ARGUING ABOUT JAVA vs C to WRITE SYSTEM SOFTWARE.

    Use C for _system_ drivers.(linux)
    User Java for the web. (apache, jserv, ejboss)

    I would like to see more interfaces in Linux done under java there are many good tools out there.
    BTW I think "jikes" the compiler from IBM is the best around (the fastest by a lot) and is the one I use day in day out. On the gui side there is freebongo (www.freebongo.org) done by the original graphic java guy (van hoff) it is kick ass, very professional. There is the blackdown 1.2 on linux wich is the one I will stick to. Component server and web integration is done with EJBoss and Jserv. (www.ejboss.org) (www.apache.org).

    BOTTOM LINE: linux and java have a great future together. The industry has already adopted java, linux is coming real fast. Portability will do MS in.

  • Try ibm's jikes compiler - it's incredibly fast.

    And the first tests using the ibm jvm w/jit is that it's more than 30x faster than blackdown's jdk117_v3 w/o the jit.

    It works for me. My favorite java development platform just got better. Thanks, ibm.
  • Yeah sure, OS/2 is at 1.1.8 and even faster.
    But the 1.1.6 implementation wasn't exactly crawling either! It was already at the speed where a good P120 wouldn't have too many complaints about Java app speed.
  • Blackdown just finished porting Java 1.2 to Linux. Now we get the momentous news that we have 1.1.6 from IBM?

  • "!ErrorBookmarkNotDefined" raised a couple of issues that I think need to be addressed, but haven't been yet.

    "I'm dubious about the need for another JDK 1.1.x port."

    Let's look at who released this JVM... I.B.M. As we all know the 'B' stands for Business. In the business world no one has applications ready for Java2 yet, and the companies who are deploying mission critical apps want the maturity of the 11 systems. IBM didn't release this for philantropic reasons, rather they have several products [ibm.com] in the market today and announced for tommorow that require a good, fast, stable, serviceable JVM under them. This same thought path can explain the release of the "IBM Developer Kit for Windows(R), Java(TM) Technology Edition". Why else would big blue want to make a JVM for the M$ desktop? Because they needed one to support their products. They couldn't use M$'s because it wasn't compliant and therefore broke a lot of stuff. They couldn't rely on the Sun reference implmentaion because it simply can't handle "enterprise" level usage. (try to keep an RMI based server up under load for more than a few hours on Sun's 117B... I dare you. Then try it on IBM's JVM.)

    IBM is the biggest player in the industry, having created dozens of different architectures over they years, and they still support a very wide variety of those hardware platforms. They have customers who are still DEPLOYED on System 34s (1960s technology) which will become permanently obsolete [year2000.com] in about 200 days. Faced with this historical dead weight IBM is the corporation with the most to gain from leveraging Java's crossplatform nature.

    "Particularly one with such a low version number."

    IBM's unix VMs have typically skipped the odd release numbers... any AIX Java user will confirm that they went from a112 to a114 to a116... each of which had at least four refreshes over it's life via IBM's PTF mechanism. I doubt a118 is finished yet, so 116 is the current IBM Unix JVM.

    Three other points raised by !ErrorBookmarkNotDefined were:
    1. automatic non-classpath byte code verification.
    2. RMI that doesn't hog port 1099, and can be tweaked with a security policy
    3. speaking of security, the boolean sandbox model in 1.1.x is either too restrictive or too liberal for e-ware. JDK 1.2's fine-grained policies are needed to do anything non-trivial in Linux java.
    Well, the answers to those are
    1. use -verify... if you don't want to forget it "fix" the wrapper script.
    2. Moving the RMI registry off port 1099 is trivial so don't give me that crap.
    3. the boolean sandbox isn't meant for real security... neither is the fine grained headache that is introduced by Java2. Real security needs to take place at the application layer as well as at the comm layer. Applications [ibm.com] which need security, or other complex issues such as transactional processing, or distributed process/thread scopes will extend RMI, or reimplement it.

    "IXX" said "Its funny that I see info on IBM software releases here before I see them on the IBM internal web site. :)" yup that's an understatement... someone inside the JTC laughed when I asked if the rummors were true, just a week ago. I get the feeling this was kept hush-hush and done quickly. Ixx continued "When I can start developing for Linux (at IBM) instead of just using it as a desk top I will be very happy :)" to which an AC responded... "Just grab the source of the project you're working on and start porting it in your free time. If you can get a working prototype, your managers will love you for it." Umm... don't be so sure. sigh.

    I don't speak for IBM,
    IBM doesn't speak for me.
    It's better that way.
  • Well I doubt we'll ever see Java as the implementation language of choice for time critical functions like rendering scenes on the fly, but it could have its uses in games. Gamasutra has a lengthy article entitled "Dirty Java: Using the Java Native Interface Within Games", here [gamasutra.com] which gives a rundown of some of the uses that Java + compiled native code is being put to and the pros and cons. Kinda interesting I thought.

    Straying way off topic theres an intriguing Linux comment from one of the Heretic II developers over on the site too. Heretic Linux comment [gamasutra.com]
  • Okay, so a bunch of questions

    1. Is this IBM's in house code or a port of the JDK? They say in the FAQ they don't have a right to the code, but blackdown's JDK license let's them release diffs, did they get a better license than IBM? :)

    2. I wonder if IBM and blackdown might benefit from joining forces here. Blackdown's done a lot of work in getting JDK 117 and 12 up and running under linux. I'm sure a couple full time IBM engineers could finish the job effectively and quickly.

    My only true complain about java under linux (I develop java under linux "grad student full time" for my research) is that the time from sun announcement to linux usable port is rather slow. This is a lousy artifact of Sun's business model. Hopefully IBM joining into the fray, Transvirtual (sp?) making their clean room implementation more usable all the time and blackdown folding their changes into the main source tree will change this ... (as well as the gnu supported classpath and japhar, but those aren't contenders just yet).
  • | Just another reason why developing Linux apps in
    | Java is a good idea.

    Not because of this particular software package. From what I can tell from the site, it's Intel only. Intel-only Linux application software is a *bad* thing.

  • In compiling code on my machine, the IBM jdk is half the speed, i.e. twice as slow as blackdown 1.2 pre 2.

    The IBM JDK is fast at starting however....
  • Have you seen Magician? They build a OpenGl canvas so you don't need glut or gltk. You write your application framing and dialogs in Java and interface to the native OpenGL system via the Canvas. Why do you think Micros~1 was so bent on NOT supporting JNI? Platform neutrality is what I think.

  • Nice, but is it what we need? I'd like 1.2 (Score:3)

    by gavinhall ( 33 ) on Saturday June 12, 1999 @05:36AM (#1853682)
    Posted by !ErrorBookmarkNotDefined:

    I was glad to see yet another 1.1.x port to Linux. However, I wonder how useful this port can be. Consider that it lacks byte coder verification of the class files found in the CLASSPATH. (Note: I assume IBM of all companies followed the white paper on this and didn't change the primordial class loader.)

    Consider that it lacks the fine-grained policy provided by JDK 1.2. Consider that the classloaders are limited to URL and sandbox models. Consider that it lacks good JFC support. (Yes, yes, I realize you can just path the swingall.jar libs, but the speed is terrible, even if the bug pathes from Sun precede those of JDK 1.2.2; other people might disagree, but I don't find non-optimized Swing a viable alternative. Plus, there's no Java accessibility, glasgow, etc.)

    This is not to fault IBM; they've done wonderful work with Java, and appear to be serious about linux as well. Instead, these are just limitations in the 1.1.x VM--among many others.

    So, while I'm glad to hear of IBM's efforts, I'm dubious about the need for another JDK 1.1.x port. Particularly one with such a low version number. There are a number of critical bugs in JDK 1.1.6 (e.g., the "contains()" and "inside()" snafu fixed only in 1.2 and 1.1.8), that are likely repeated in this port. (Perhaps I'm wrong; but then why not version the package up to 1.1.8?)

    Let's look at the field of 1.1 VM so far, according at least to FSF's tally:
    • BISS AWT [biss-net.com] - A Java framework for building graphical applications.
    • Jlint [ispras.ru] - Java program checker
    • Kaffe [transvirtual.com] - A virtual machine to run Java bytecodes. For many architectures, "just-in-time" native code generation is supported.
    • Kiev [forestro.com] - A compiler that extends and unites the Java and Prolog languages.
    • Classpath [slashdot.org] - Essential libraries for supporting the Java language.
    • SableCC [mcgill.ca] - An object-oriented framework that generates compilers (and interpreters) in the Java language.


    All of these distros are capable 1.1.x compliant VMs. I was glad to see them when they arrived. But they all lack many of the key improvements in the 1.2 VM:

    1. automatic non-classpath byte code verification.
    2. RMI that doesn't hog port 1099, and can be tweaked with a security policy
    3. speaking of security, the boolean sandbox model in 1.1.x is either too restrictive or too liberal for e-ware. JDK 1.2's fine-grained policies are needed to do anything non-trivial in Linux java
    4. I could list this stuff all day


    It strikes me that without javax support for crypto, security, non-port-specific RMI, and other "enterprise enabling" packages, the 1.1.x ports are not that helpful. Of course, someone wanting to make a nice GUI or web client can use the IBM port. (For that matter, why not use blackdown's more mature port?)

    So, while we should welcome this distro, I think the Linux Java community needs to press for a good 1.2 VM. (I.e., our debug and testing cycles should be applied to, say jdk1.2pre2 from blackdown, so that thread safety, RMI, drag and drop and other "e-features" start working right.) It's great the IBM has a 1.1.x distro, but it comes kinda late.

    I'm sure IBM could amaze us all with a good JDK 1.2 linux port. It would be nice to see the speed that the 2.2 kernel has over Windows fully exploited in a good port.

    [Note: Again, I'm not looking a gift horse in the mouth; it's just that we need something stronger than yet another horse.]

    -----------------------------
    Computers are useless. They can only give answers.
  • Posted by lamikr:

    >o, while I'm glad to hear of IBM's efforts,
    >I'm dubious about the need for another JDK 1.1.x
    >port. Particularly one with such a low version
    >number. There are a number of critical bugs in
    >JDK 1.1.6 (e.g., the "contains()" and "inside()"
    >snafu fixed only in 1.2 and 1.1.8), that are
    >likely repeated in this port. (Perhaps I'm wrong;
    >but then why not version the package up to
    >1.1.8?)

    With OS/2 I'm officially still using Netscape
    4.04 but it still includes security etc. fixes included to Netscape releases with little
    higher version-number like (4.05-4.07)...

    Maybe same will also happen with JDK 1.16 for
    linux when IBM will release final version of it.

    Mika

    Btw. of friday IBM also released JDK 1.18
    beta for OS/2 and this package contains
    some stuff from JDK 1.2...
    (And with this package comes with some other
    extra goodies like ICAT source debugger
    which is heavily used also by device-driver developers...)
    devi
  • Posted by !ErrorBookmarkNotDefined:

    [Somewhat offtopic]: Hey, these articles aren't scored. Kinda strange for /.
    Perhaps there are not too many moderators for Java issues? I had to adjust visibility to read the threads--many of which rise above the pissing war of language vs. language, Java vs. C, Java vs. Godzilla, etc..


    -----------------------------
    Computers are useless. They can only give answers.
  • Here ya' go. Times for a simple embedded loop.

    blackdown
    jdk117_v3 - 33255ms

    ibm 1.1.6 - 32835ms

    ibm 1.1.6/jit - 975ms

  • I wasn't complaining about compile times. I was using it as a very simple benchmark.

    FWIW, JDK1.2pre2 is still faster than IBM's JDK doing non-GUI stuff other than compiling.

    I know about Jikes, but don't use it.
  • Here at K-State they use it for their Programming Fundamentals (CIS200) class. Its worthless and I depise it. I learn a hell of a lot more taking C for Engineers (CIS209). It was more intuitive and damn sure more useful. Where would I use a little java? Sprucing up my webpage would be my *only* consideration. Period.

    Justin
    --
  • java fully has my acceptance open source or not.

  • actually java code is virtually the same as C. the differences if any, are minor at best. (this coming from a die hard C programmer). i never thought i'd touch java but when i looked at it my C programs could be ported over with minimal effort. that, and it runs machine independent over the web.
  • this may be a non issue, but it would be kewl to have OS/CPU independent fonts available at runtime for java apps. i know 1.2 does that but 1.1 is sorely lacking. a quick hack is to check for the OS and change the fonts accordingly..but thats just a messy hack. Between irix and doze/linux/OS/2 systems this is a real pain.
  • Its funny that I see info on IBM software releases here before I see them on the IBM internal web site. :) When I can start developing for Linux (at IBM) instead of just using it as a desk top I will be very happy :)
  • The JVM on OS/2 is very speedy and I was getting quite concerned since IBM brought that speed to Win32. Now its arriving on Linux and with VisualAge for Java arriving soon too, I can be fat, dumb, and happy in Linux just as I've been for years on OS/2. This is great. Now if we can just start seeing some of those new configuration utils written in Java, we'll be style'n. I'd be more willing to lend a hand with a Java-based Apache config util then a pearl or tcl/tk one. Just my opinion.
    Again, this is great. The download is complete so off to boot into Linux to give it a whirl. Be back soon.....(funky way to download, no single file and you have to click through the license on each file)
  • If you find current implementations of Java somewhat sluggish, you won't believe the speed of this baby. On my P200/64MB running OS/2 Warp 4, this thing just blows everything away. One of my friends found the speed to be similar (or better) compared to Microsofts JVM for NT on a P2-350/128. Add a 100% pure implementation to that, and you have a killer. This JVM is what made me change my mind about coding for Java, and it will probably affect you too.

    Oh yeah, this thing is fast.
  • Financial institutes and banks have good reasons to be deeply involved with Java.

    The average gameplayer? Mmm .. Maybe not.

  • Re:Open Source? (Score:1)

    by Anonymous Coward
    Check on Cygnus' native Java compiler. It's totally GPL'ed.

    http://sourceware.cygnus.com/java [cygnus.com]
  • Are you kidding? Compiled java is one of the best things to have happened to the computing industry ever! It takes its style from the greatest programming language of all time - C, adds OO onto it beautifully, and does all the groovy ada95-like stuff (exception handling, memory nanny, etc).

    As for interpreted java... well... it's a different field I suppose. But would stuff like webmeasure be possible without it? Also, little apps done well in java are really nice. The problem is that people have always thought of java as just a beefy web scripting language... which is a bad thing.

    - C
  • IBMs JVM w/JIT blew the TowerJ code away in one of the recent benchmarks (Volano?). Bringing the performance of the OS/2 JVM to Linux should be a boon to Java on Linux. It would be nice if it was OSS but Micros~1 is still a threat and can not be allowed an inch toward damaging Java. Give'm an inch and they'll take a yard, they've show that already. I think Java will eventually be opened up but it has to become more widespread. It is still a niche on the client even though it could replace VisualBasic today. What does the GPL give you if you can look at Sun's source now and use IBM's JVM with your product? Let's see how responsive IBM is to bug fixes first, they have been very good with fixing bugs found in the OS/2 JVMs.

    LessTif just finished downloading so back to seeing how fast JStreet Mailer is on this new JVM. MoneyDance gets tested after that.....
  • Do what I did.. (On the South AIX desk) Find every single unused PC in sight and install linux on it. People will get used to it, fast. Then they will covet it. It grows.

    --
    blue
  • "Oh yeah, this thing is fast. "

    Sounds to me like you are thinking about the latest JVM released for Win and OS/2?

    This baby is the somewhat older 1.1.6 implementation. Needs Motif as well.
  • It won't be too long until games are being written in Visual Basic using DirectX7/8, with all of the numerically intensive geometry transformation, lighting, and rasterization handled by hardware, leaving the developer to deal with scripting the story line, and art, and not writing game engines.

    This, of course, will be known as the death of games. Sure there may be a brief period that produces a couple of good games, but it won't last once the Nintendo Effect sets in. ( Nintendo Effect: Hey! This new game, Super Max Attack IV is just like the old game Mega Blaster Man 8 except the bad guys are blue instead of green and the players gun goes bleep insead of bloop.)

  • I've got it running. Since I'm on OpenLinux v2.2 I needed to download LessTif and install it. To do that you have to be sure it is the libc2 version and I only found a binary tar-gzip distribution. unpacked to /usr resulting in /usr/lesstif. Edited /etc/ld.?.conf and added /usr/lesstif/lib to the end of the file and ran ldconfig. That is it. Runs pretty fast and only feels slightly slower then the OS/2 JVM v1.1.7a. I've had a problem getting JStreet over since hpfs support does not take hpfs's filename correctly. Took me some time to realize that was the reason my user names and mail stores weren't found.
    I don't have pcmcia support in v2.2 yet since I installed using the GUI install and it says I 'can't get there from here'. No benchmarking yet
    because of being Inet-less.

    Love to see some benchmarking/comparisons posted...
  • All of these distros are capable 1.1.x compliant VMs.

    Um, get a clue. Only Kaffe is actually a full VM. Classpath isn't a VM and isn't compliant to anything because it isn't ready yet. JLint is clearly not a VM.

    BTW, the dmoz [dmoz.org] directory has a more comprehensive listing of open source Java stuff than the FSF:

    http://dmoz.org/C omputers/Programming/Languages/Java/Open_Source/ [dmoz.org] (86 links - and that's just the open source links)

  • Me thinks IBM is feeling the waters. The announcement about VA for Java on Linux was a technical preview and will only become a product if demand dictates. Bringing it up to 1.1.7 shouldn't be too far behind. Regarding Motif, LessTif 0.88.1 works too. It would be nice if it was OSS since it is really a OS runtime env. We will have to see what Sun does in the next couple of years.
    I did find a bug in that the message content window of the JStreet Mailer (www.Innoval.com) has blacked out text with this JVM. By selecting the text it becomes visible. It crashed once on a dialog too. Very good start though considering its speed and knowing the quality of the OS/2 JVM.
  • i find it disappointing that the licence prohibits
    mirror sites from making this available. :-(

    perhaps ibm don't quite understand what opensource
    is all about, when they're still talking about only allowing people to make `backup copies' of software they want feedback and usage on.

    -jason
  • The company I work for is 100% Java oriented,
    and we do quite well out there in the real
    world.
  • I'm pretty sure IBM do understand what OpenSource is about. However, this package can never be opensource, since it's based heavily on Sun's code, for which IBM had to sign a licence.

    As for mirroring, I think you'll find that the main issue is that IBM wants to know how many people are interested, to help them gauge the size of the market. I can't see anything wrong with that -- I'd be interested to know myself.

    I've just downloaded my copy now, and IBM seem to have enough bandwidth to keep up with demand :).

Slashdot Top Deals

For God's sake, stop researching for a while and begin to think!

Close