Java2 SDK v. 1.4 Released

pangloss writes: "Yay: XML, built-in Perl-ish regex, jdbc 3.0, asserts, IPv6, lots of other goodies. Release notes and incompatibilities. And I think this means I can use my wheel-mouse in NetBeans without that extra module ;) Download it here." WilsonSD adds: "There are many cool new features including a New I/O package, an Assert Facility and enhanced performance." Some other random Java notes: O'Reilly has an essay about why you won't see any open source J2EE implementations, and Kodak has filed a patent-infringement claim against Sun regarding Java.

Comments

It's a big step up, but there is still distance

[btempleton]

I've been working on a server that takes a lot of connections in Java, and you can finally do it with the support of "select" in Java 1.4. But no support for SSL. I undertand why it happens, but this "we'll get around to doing the security later" is why we don't have a lot of security.

Java will always present a dilemma. With the push for portability, you often have to wait for the platform itself to support things like this (select) or kludge it in very non portable ways. But that portability leaves the system behind which hurts it in competition with other systems more tolerant of innovation and the fracturing it brings.

A good philosophy would be to rule that every time a system library or feature needs to do something that an ordinary user can't do, they don't just build it in, they make a way that an ordinary user could write it. That paves the way for more innovation.

Re:It's a big step up, but there is still distance

[Anonymous Coward]

SSL at the Java level IMHO is stupid. Set up secure tunnelling whenever the data leaves your trusted zone, and have it be done in hardware or optimized (C) software.

Re:It's a big step up, but there is still distance

[btempleton]

You assume you control the zone of the connecting client. If you want people with a browser to connect to you securely from out there anywhere, SSL is your choice.

Re:It's a big step up, but there is still distance

[Wraithlyn]

"I've been working on a server that takes a lot of connections in Java, and you can finally do it with the support of "select" in Java 1.4."

I've been using NBIO (Non-Blocking IO) for quite some time, getting very good results. Been waiting for Java 1.4 to go final so I can start working with java.nio

Re:It's a big step up, but there is still distance

[blamanj]

It's worth noting that the Berkeley NBIO package was part of the research that led to the 1.4 java.nio package.

Also, using their package, and a very interesting partitioned, event-based coding style, they wrote, in Java, a web server that out performs Apache (written in C).

Re:It's a big step up, but there is still distance

[ggruschow]

You should probably hold up for a while. The Selector still has a few serious bugs, some of which they've acknowledged for >10 months now without fixing.

Don't get me wrong, I'm using the NIO libraries anyway because my company's system needs them, but I was running into a new bug a week for a while. I was hoping they'd have fixed at least the major ones by the time it became official. The workarounds for some of them are just awful.

My major sore points:

• The Selector only supports 63 channels on Win32 (that'll help your scalability :)
• The Selector doesn't work the same across platforms. On Win32 works like WSAAsyncSelect() (only reports channels when something was blocked before), it acts like select() (reports channels whenever they match the requested ops). (Damnit. Write-Once-Run-Just-On-UNIX?)
• Selector.wakeup() breaks after 4096 calls on Linux (Guess I just won't make any new connections :)

hope swing is faster now / I prefer Ruby over Java

[raistlinthegreat]

I hope that Swing is faster now under linux. I played with netbeans and jedit and although there are really good they are not very fast on linux. on a slower windows PC jedit stated much faster than on my linux box.
a few days ago I played a little with Ruby (the coolest language available IMHO) and Gtk+. although the bindings are not yet finished they work very well under Linux and this is much faster than Java/Swing.
Maybe Ruby is the future. At least I hope so. If Ruby gets more stable modules Ruby can be the Number One OOP language. It is cleaner than Perl or Java, the Programms are shorter, the Language is more intuitive and.... and.... and. This is only my humble opinion. See for yourself if you like ruby. check out http://www.ruby-lang.org

Re:Yah, will this be stable on Linux?

[RMSIsAnIdiot]

Of course it will be stable. Everything is stable under L---

kernel: VFS: Disk change detected on device fd(2,0)
kernel: end_request: I/O error, dev 02:00 (floppy), sector 0
kernel: VFS: Disk change detected on device fd(2,0)
kernel: end_request: I/O error, dev 02:00 (floppy), sector 0
kernel: Incorrect segment count at 0xc01781d6nr_segments is 15
kernel: counted segments is 17
kernel: Flags 0 0
kernel: Segment 0xc37ace40, blocks 4, addr 0x4bd1fff
kernel: Segment 0xc37acea0, blocks 4, addr 0x4bd27ff
kernel: Segment 0xc37ac780, blocks 4, addr 0x38abfff
kernel: Segment 0xc37acde0, blocks 4, addr 0x38ac7ff
kernel: Segment 0xc3809120, blocks 4, addr 0x1e77fff
kernel: Segment 0xc0e0fda0, blocks 4, addr 0x199c7ff
kernel: Segment 0xc3871ec0, blocks 4, addr 0x6fe2fff
kernel: Segment 0xc3871f20, blocks 4, addr 0x6fe37ff
kernel: Segment 0xc3871e00, blocks 4, addr 0x498afff
kernel: Segment 0xc0e0fec0, blocks 4, addr 0x18aefff
kernel: Segment 0xc3871c80, blocks 4, addr 0x41b6fff
kernel: Segment 0xc3871ce0, blocks 4, addr 0x41b77ff
kernel: Segment 0xc3871b60, blocks 4, addr 0x1d06fff
kernel: Segment 0xc0e0fe60, blocks 4, addr 0x18af7ff
kernel: Segment 0xc38719e0, blocks 4, addr 0x4c3ffff
kernel: Segment 0xc3871aa0, blocks 4, addr 0x4c407ff
kernel: Segment 0xc0e0ff20, blocks 4, addr 0x18ae7ff
kernel: Segment 0xc3871800, blocks 4, addr 0x4cb0fff
kernel: Segment 0xc3871860, blocks 4, addr 0x4cb17ff
kernel: Segment 0xc3871140, blocks 4, addr 0x4b8ffff
kernel: Segment 0xc38717a0, blocks 4, addr 0x4b907ff
kernel: Segment 0xc0e150c0, blocks 4, addr 0x196d7ff
kernel: Kernel panic: Ththththaats all folks. Too dangerous to continue.
kernel:
kernel: rq->cmd=1, rq->sector=73208, rq->nr_sectors=8
kernel: kernel BUG at pktcdvd.c:1046!
kernel: invalid operand: 0000
kernel: CPU: 0
kernel: EIP: 0010:[serial:__insmod_serial_S.bss_L3392+238989/37 155027]
kernel: EFLAGS: 00010086
kernel: eax: 0000001e ebx: c1273840 ecx: c4676000 edx: c020f384
kernel: esi: c208f09c edi: c208f074 ebp: c7f9d760 esp: c6fc3f88
kernel: ds: 0018 es: 0018 ss: 0018
kernel: Process pktcdvd0 (pid: 1303, stackpage=c6fc3000)
kernel: Stack: c8861c10 c8861d2f 00000416 c6fc2000 c208f09c c208f074 c208f000 c7f9d86c
kernel: c8860076 c208f074 00000f00 c762df1c c208f000 00000b00 c6fc3fe0 c208f008
kernel: c208f10c c7f9d778 00000000 c6fc2000 00000000 00000000 00000000 c6fc2000
kernel: Call Trace: [serial:__insmod_serial_S.bss_L3392+245712/3714830 4] [serial:__insmod_serial_S.bss_L3392+245999/3714801 7] [serial:__insmod_serial_S.bss_L3392+238646/3715537 0] [kernel_thread+40/56]
kernel:
kernel: Code: 0f 0b 83 c4 0c b8 06 00 00 00 0f ab 45 48 19 c0 85 c0 75 24

Summary of new features

[LarsWestergren]

A summary of all the new features is available here:
http://java.sun.com/j2se/1.4/docs/relnotes/feature s.html

Articles about the news APIs and how to use them available here:
http://java.sun.com/j2se/1.4/articles.html

Re:Summary of new features

[khuber]

Hilarious. Basically a reader-friendly translation of "We hardcoded what was previously a free parameter because people were too st00pid to use it properly."

And probablePrime() was there before, they just changed it from private to public.

-Kevin

Coincidence?

[Yakman]

I find it funny that the post mentions a lawsuit from Kodak and the O'Reilly essay about Java says:

[...] At the same time, Sun loves to have Kodak moments with some parts of the open source community -- most notably, Apache -- who increasingly feel used and abused [...]

(Emphasis mine) Heh.

Asserts

[Malc]

This makes me so happy. Coming from C++, I really really missed assertions. They give me much more confidence in my code. Some people seem to have trouble using them, but after a while they can become second nature. In fact, one can come to rely very heavily on them.

Re:Asserts

[utdpenguin]

Assertions can be dangerous. Many people make confident assertions about subjects concerning which they have little knowldge. For example, me talking about coding. :)

Re:Asserts

[Anonymous Coward]

Yes. Asserts are easy to use and have a huge payoff. They are especially good at catching nasty subtle bugs.

They are easy; they catch bugs that might not otherwise ever be noticed (or noticed only as some pervasive flakiness); they save you a lot of time that would otherwise be spent debuging; they are good documentation; they don't cost anything in non-debug builds; they save you from a lot of pain. Despite all this, I haven't yet successfully convinced a single person who is unfamiliar with them of their value. Christ I get dismayed sometimes.

This interview with FreeBSD kernel hacker Matt Dillon has some interesting things to say about assertions ("it greatly contributed to our famed stability in 4.0 and later releases [of the FreeBSD kernel]").

Java finally has them? Cool. What year is it again? 2002? Jeez ...

Re:Asserts

[rbeattie]

I'm one of those people who really don't get assertions either. But regardless, from what I've read in order to use them, you have add a "-source 1.4" flag on your javac command in order to compile which I think is weird... I mean, if you're going to add a feature, you should add it and not start messing around with flags.

-Russ

Re:Asserts

[Steveftoth]

Not only that, but if you actually want the assertation to be checked during runtime you have to enable them when the VM is started.

Though this is kinda cool, cause then you can use asserations as a 'runtime' debug switch where you put debug statements into assertations and then turn them on when you discover a bug.

J2EE openness

[prockcore]

Well that kind of gives an answer to every javahead that says "Why bother with Mono, when you have J2EE? What a waste of time."

Mono is at least opensource... can you say the same for J2EE? Will you ever be able to say the same?

Re:J2EE openness

[customiser]

Sun certification does not have any effect on JBoss's functionality. As I see it, it is just a marketing matter, being able to say "This is a Sun certified J2EE App Server", so that whoever makes the decision on using it (and mainly commercial organisations) can be confident that it really does what it is supposed to do.

btw, as it is a server, it does not run on the client.

Genericity?

[hephro]

And still no generic data structures (a.k.a. templates in the C++ world)... all those explicit downcasts from Object hurt and need to be optimized away by the JIT...

-Hein

Re:Genericity?

[Westley]

The proposed generics proposal are all compiled into the bytecode anyway, I believe. There's a prototype compiler (which generates bytecode which can be used with the normal JRE) available here.

Jon

Re:Genericity?

[srichman]

1.5

Re:Genericity?

[gatesh8r]

From the article:

Bracha pointed out that one problem with C++ templates is code bloat. For every List<T> of type T, C++ generates a separate class.

I'm glad someone posted this link. This is EXACTLY the problem with C++. It isn't in the class structures (they are almost exact to C's structs) but the fact that the templates have to generate seperate class structures to make sure that each one doesn't conflict.

This I have seen working with C++: A matter of a template has done almost a 25% - 80% reduction in my executable size. I'm not kidding -- from 1 MB to 200 kB! This was the case because I made two instances of the same class (cut 'n paste) because I really didn't need all the generics involved in the program. While typing all that stuff then doing a cut-n-paste was slow to do, the executable size was worth it.

Anyway, if you code in C++, the STL is decent FWIW. I wouldn't lean on it though if you're going to do major embedded coding... but get the String class; it's great.

Re:Genericity?

[Twylite]

Java will support parametised types, not templates. C++'s solution is a hack, and most people know that. True language (and VM) support for parametised types will not involve any binary bloat. In a way you can think of it as having a "object type" attribute and doing an "instanceof" to check that the type is right.

Re:Genericity?

[Rogerborg]

• still no generic data structures

And I was recently delivered C++ code that used a class with a private constructor and const static instances, instead of good old fashioned enums (goodbye switch, hello if/else/if/else/if/else...). I initially assumed that it was a paranoid developer who'd read Effective C++ and didn't want to be passing uninitialised enums around. Turns out that it was written by a Java programmer who didn't know what an enum was.

Java isn't just missing enums, it's beginning to remove them from the developer's toolkit, and that can't be a good thing.

Re:Genericity?

[khuber]

1.5 [javaworld.com]

The 1.5 Java generics are merely syntactic sugar. Since the casts still occur, there is no performance benefit, only cleaner source code. There is no way to parameterize primitive types like int.

I do not consider the proposed generics to be adequate. You are still better off, performance wise, generating your own custom collections for specific types. This is what GNU trove does.

-Kevin

Re:Genericity?

[khuber]

Cleaner source code!? How about compile-time type safety?

Yes, obviously, but I've never had much of an issue with it. Have you? That seems like more of a purist argument to me. You can always subclass your own collections for that. Sure it would be good and more convenient, and I'm not disagreeing that compile-time safety should be there, but an invalid Java cast will throw a ClassCastException. It's not like casting a C pointer where you will probably segfault.

My issue is just that generics only address part of the issue. Casting and using wrapper classes like Integer is slow. Once you have fixed the class types at compile time via generics, why must you take the cast hit at runtime? It's a hack. If you build your own collections for specific types, you have compile time type safety -and- performance.

-Kevin

Re:Genericity?

[Jagasian]

Some functional languages have had strong higher-order (i.e. polymorphic/generic/parametric) types since the late 60s. If course, in an Object-Oriented world, your gotta be an Object-Oriented girl. Object-Orientation reminds me of the SUV. It's so 1990s sheek!

Re:Genericity?

[Elbows]

On large projects it can be a big issue. Someone
puts an instance of class X in a vector that's
really meant to be holding class Y. No exception
is thrown until you access the bad data and try
to cast it to a Y, at which point there's no good
way to find out where the error occurred.

Now you have to dig through 100k lines of code
to find the mistake. With generics it's caught at
compile-time.

That said, generics should be implemented so as
to give you performance, too. If you know
everything in a collection is an X, there's no
reason to do any casting at all.

64 bit - wow!

[Florian Weimer]

The 4 GB address space limit has become a severe limit on Java bloat. It's good to see that Sun finally addresses this problem.

Re:64 bit - wow!

[ckd]

The 4 GB address space limit has become a severe limit on Java bloat. It's good to see that Sun finally addresses this problem.

Compaq already shipped a 64-bit JVM for their Alpha systems (running Tru64, OpenVMS, or something called Linux) a while back.

J2SDK on Win98 and Linux

[Larkfellow]

I have a dual boot machine at home, one partition Windows 98, the other Slackware 8.0. Today I downloaded and installed the new Java SDK 1.4.0-rc on both systems. And while I think that some iscolated windows difficulties causes my oppinion to be rather biased, I found the install much easier going on Linux.

However I will note that, while the Java Web Start was installed on Windows, I didn't find any version of it for linux. And the downside to the Web Start I found is that it constantly wanted to download and install a new version of Java Runtime Environment 1.3.x everytime I lauched an application. And then after the download, and installed, I'm prompted to reboot the computer. After rebooting and trying to launch again, it again starts to download JRE 1.3.x and through the whole cycle all over again.

As well, with my windows install I found I was constantly having difficulties getting it to use the default classpath (ie, no environment variable set for CLASSPATH). I ended up having to resort to specifying the classpath at the command line. And no matter how much I tried, I could no manage to get Swing to work properly.

However on the other hand, the Linux install was rather straight forward, with a few simple steps: Change download to executable; Run it; move the extracted directory to a shared path (as su); add the java/bin directory to the search path; and finally add the java/man directory to the search paths for man.

The windows installer was straight forward, though the above problems still hampered me.

Re:J2SDK on Win98 and Linux

[Westley]

No, all is unlikely to be hunky-dory if you set the classpath environment variable. This seems to cause more problems than anything else to newcomers. It's much easier to avoid using a classpath environment variable in the first place - the default is fine for most things, and the extensions mechanism really helps. I've got short essays about both on the Java bit of my site.

Jon

Re:Java2 ?

[LarsWestergren]

Hmm, if SDK v1.2 became "Java2", shouldn't v1.4 be called "Java4"? Oh I see, "Java2" was a "marketing trick" used in the tough year 1997 and the name has stuck.

Perhaps,but remember that the change from 1.1 to 1.2 contained some major rehaul of APIs.

In this 1.4 release Sun kept a carefully budgeted set of new features and emphasized quality and performance, like they said they would. You can expect more changes in 1.5 (codename Tiger) which is planned to come some time around the middle of 2003. If that will be Java3, who knows, they have just started working on it.

I've been using this for a while

[autopr0n]

Autopr0n.com actualy runs using the first beta of JDK1.4. I needed the new ImageIO libraries for the, um, 'site previewer :P' (and the regexs made some of the parsing I'm doing a lot easier :). I tried 1.4b3 but it was far more unstable, and my regexs broke.

Definetly cool that the stable version is out, I'll have to upgrade at some point, when I have time.

Hopefully they'll implement this at Topcoder.com too, so I don't have to keep using the old docs in the compos :P

Re:I've been using this for a while

[Bodrius]

At least they don't include the 300 Javascript Pop-Up Ads of Death.

Re:I've been using this for a while

[flacco]

Autopr0n.com actualy runs using the first beta of JDK1.4. I needed the new ImageIO libraries for the, um, 'site previewer :P' (and the regexs made some of the parsing I'm doing a lot easier :). I tried 1.4b3 but it was far more unstable, and my regexs broke.

It must be hell on productivity working on Autopr0n, what with the spank sessions every thirty minutes or so.

At least Sun is brutally honest.

[Marsh Jedi]

Jesus, Sun's PR corps must have flipped their collective shit when they read Karen Tegan's remark. While in general I find that kind of bluntness refreshing, a director of Platform Compatibility spouting off and essentially saying, "Well, that sucks for you, but we make money off of compatibility testing. We give everything else away for free, so cut us a break." is really a testament to the Sun Micro (brutally) plain-talking attitude.

Deeper, though, I think, is the need to rein in Java a bit....It has achieved ubiquitousness, and I think Sun knows it. Watch. If .NET takes off, they will loosen up to benefit from a little more old-fashioned agrarian innovation and buzz.

To lazy to write it yourself?

[autopr0n]

I guess it would be nice to have SSL built into java, but you could write it yourself, it's not like it's a fundementl language flaw. The specs are out there, and the java Crypto API would probably help you out quite a bit.

Re:To lazy to write it yourself?

[spiro_killglance]

SSL is in java, at least at the https: level, and
has been since JDK 1.3, you just need some additional SUN libraries.

download and add, jsse.jar, jcert.jar, jnet.jar.

And you can open URLs like

URL u = new URL("https://where.ever.com/file");

and can make connections on them, just the same
as you could any other URL.

there are plenty of OSS java implementations

[autopr0n]

see subject

AUTOEXEC.BAT

[autopr0n]

Throw the classpath in your AUTOEXEC.BAT file on the C: drive on your win98 machine. Should work.

My take on JDK 1.4

[burtonator]

OK.. I am a Java fan... (recently this has been changing though.)

I have mixed feelings with JDK 1.4.

The JPDA (debugging) support in 1.4 is vastly improved. You can now redefine classes in a running virtual machine. This is really cool and I have written an Ant 'Redefine task to take advantage of this.

The assert facility is OK.... i don't like the fact that they added an Assert keyword but I don't get to make the decisions.

There is also some controversy.

The JSPA agreement that one has to sign to participate in the JCP is WAY too restrictive for Open Source developers. The Apache Software Foundation has a good document where they drawn the line in the sand on their participation.

The Log4J people are upset because there is now a 'stanard' Java package for logging. IMO the 'standard' package is inferior to Log4J in many situations.

The regexp package is not all it is cracked up to be either. I would recommend Jakarta ORO or Jakarta Regexp.

As far as that... it runs GREAT on Linux. Probably the most SOLID VM I have ever run.

They did break some stuff with legacy code. If you ever named a class 'URI' your code will now fail to compile because they put this class in the java.net package which everyone imports anyway.

As far as C# vs .Java. I am really impressed with the CLR/CLI stuff. Right now, as it stands, Java is a proprietary language. Unless we see SUN Open Source Java (or push it through a standards committee), we *may* see a JDK 1.5... but no one will use it.

Also.. check out my Reptile project. It is Java based, only requires JDK 1.2 and incorporates some really cool Java/XML stuff. :)

Re:My take on JDK 1.4

[LarsWestergren]

Thanks for a well written article.

The JSPA agreement that one has to sign to participate in the JCP [jcp.org] is WAY too restrictive for Open Source developers. The Apache Software Foundation has a good document where they drawn the line in the sand [apache.org] on their participation.
[...]
As far as C# vs .Java. I am really impressed with the CLR/CLI stuff. Right now, as it stands, Java is a proprietary language. Unless we see SUN Open Source Java (or push it through a standards committee), we *may* see a JDK 1.5... but no one will use it.

Now that Microsoft is kept busy plagiarising...oops, I mean creating a competing C#, perhaps Sun wont have to fear embrace and extend tactics against Java itself, and dares to open up a bit more to open source and open standards. Maybe they will even be forced to do it to survive.

Re:My take on JDK 1.4

[mccalli]

They did break some stuff with legacy code. If you ever named a class 'URI' your code will now fail to compile because they put this class in the java.net package which everyone imports anyway.

Not really a 'breakage'. If you imported only what you needed (java.net.URL, java.net.Socket etc.) your code will continue to work. Only if you used the statement "import java.net.*" will it now fail, and that's down to the individual coder, not the JDK.

Cheers,
Ian

Re:My take on JDK 1.4

[briansmith]

As far as C# vs .Java. I am really impressed with the CLR/CLI stuff. Right now, as it stands, Java is a proprietary language. Unless we see SUN Open Source Java (or push it through a standards committee), we *may* see a JDK 1.5... but no one will use it.

I agreed with your whole post except for this statement. It is just FUD (perhaps unintentional). The fact is that CLR/CLI isn't going to kill off Java in the next 16 months, which is when I expect to see JDK 1.5 delivered. Also, there is nothing stopping open-source clean-room implementations of anything in J2SE, AFAICT. You just can't call your product "Java" and you won't be able to get your runtime certified by Sun. Similarly, I doubt that Microsoft will let somewhat create a clean-room CLR/CLI and call it "The .NET Framework", and they don't even have a process in place for certifying compatible implementations.

Re:My take on JDK 1.4

[mlinksva]

"no one will use [JDK 1.5]" is obviously an overstatement, but the sentiment agrees with my intuition: Sun needs to make concessions to the open source community (which are in its own interest!) over the next year, or a free .net implementation will start driving java out of the niches it inhabits and could inhabit in the free software world. Long term that consigns Java to being this generation's COBOL.

Re:My take on JDK 1.4

[FastT]

i don't like the fact that they added an Assert keyword...

Why? It's not like it will realistically break anything, and it's far better than the usual Assert class and associated techniques.

The Log4J people are upset because there is now a 'stanard' Java package for logging. IMO the 'standard' package is inferior to Log4J in many situations.

So use Log4J, nothing stopping you.

The regexp package is not all it is cracked up to be either. I would recommend Jakarta ORO or Jakarta Regexp.

Yeah, except that the better of the two (ORO) has inexcusable bugs, even after all the years it's been around.

Re:My take on JDK 1.4

[Speare]

They did break some stuff with legacy code. If you ever named a class 'URI' your code will now fail to compile because they put this class in the java.net package which everyone imports anyway.

If you have a foo.bar.URI class, and the core has a java.net.URI class, you can still use yours.

You can:

1. specifically name the package at the usage (messy),
2. specifically include only the java.net classes you want, one by one, rather than including java.net.* (inconvenient),
3. or just clarify which URI class should be used in the imports specifically:
   • import foo.bar.*;
     import java.net.*;
     import foo.bar.URI; /* hides java.net.URI */

Existing code does not need to be recompiled, since bytecode always explicitly names classes always, but existing code does potentially need to be fixed if recompiled, as the default results of the imports will change. This is a pretty small and common occurrence with a new API set.

Re:My take on JDK 1.4

[Richard_Davies]

Regarding the logging and regex packages: Just because a package is less functional does not mean that it is intrinsically bad. If the package is suitable for the majority of uses by the majority of developers, then it's probably OK - after all, it's easier for someone to learn a small package rather than a large one. If you require something more specific, then you are still free to use the packages you metnioned. The JDK logging and regex packages ADD choice - surely this is a "good thing"?

Please take a close look at both the openess of .NET and the multilanguage capabilities. Neither are everything they are cracked up to be. Only the CLR and a "core" set of C# classes are open - everything else (i.e. the really useful bits that everyone needs) are not. My question - do you trust Microsoft to open these up?

You mention (must have seen this somewhere before on Slashdot :-) that Java should be Open Source / Standardised. I, like many Java developers have no instrinsic problem with this. However, there is the issue of cross-platform portability:

Many people complained when Sun would release a JDK for Windows and Solaris that it didn't have one of Linux. Then they complained when a Linux JDK was created that it didn't come out at the same time. Now with Sun releasing all 3 JDK simeltaneously (and the likes of Apple and IBM not usually far behind), consider this:

How likely do you think this situation would be if the JCP (or something like it) was not in place? Do you really think you would be saying "As far as that... it runs GREAT on Linux. Probably the most SOLID VM I have ever run." if Java was Open Source?

Maybe it would be:
"Well the Linux version is pretty good - can't use the xyz library because that's Windows only and it will probably be out of beta only 6 months after the Windows version but hey - it's Open Source! That make me FEEL GOOD!"

What I would LOVE is to see Java open sources while ensuring that it remains cross-platform. While some would claim that open source would guantee that, it is not provable. Sun believes that there is too much risk. While you may not, agree with that you have Java that is:

a) free (as in beer)
b) you can read the source code the the whole API
c) you can change (but not distribute) the source
d) works on all major plaforms (including FreeBSD now BTW)

For me, and many other Java developers, these still place it far ahead of anything Microsoft is doing - and while Mono iterests me, its going to be a LONG time before it can match Java's (or even .Net's) current functionality.

Re:My take on JDK 1.4

[Kerg]

The Log4J people are upset because there is now a 'stanard' Java package for logging. IMO the 'standard' package is inferior to Log4J in many situations.

Then why on earth would they be upset? I never understood this line of thinking... if their product is better, people will use it. I for one am happy that there is finally a standardized approach to logging. If it seems it doesn't do enough to fit my requirements then I'll use some 3rd party logging package.

So they didn't clone the log4j API, big boo hoo.

Re:My take on JDK 1.4

[pmz]

import java.net.*

Yes, '*' is the lazy programmer's haven.

Java is great, but wildcard imports just invite namespace clashes and poor documenation of dependencies in the software.

Just using one import per class provides instant automatic source-code-level documentation of all the dependencies in a software project.

Re:My take on JDK 1.4

[The Mayor]

They are upset because they offered log4j to Sun as part of the JCP to include in JDK1.4....and Sun rejected it, pushing their own implementation instead. The idea of the JCP being run by the community is a crock of shite. Sun has pushed a number of inferior standards through it, most of which are used more frequently than the superior product because many programmers are simply too lazy to use the superior product.

Stuff like logging APIs have enormous added value if everyone uses the same logging mechanism. Applications using multiple libraries that use disparate logging mechanisms are a mess. The result is people create kludges to integrate all of the various logging techniques so that they output to a single place. Sun's rejection of log4j will result in this happening.

Log4j is really better than Sun's. Oh well.

And nobody is upset that Sun didn't clone log4j. Everyone is upset because Sun didn't *take* log4j lock-stock-and-barrell. Apache/IBM offered it for the taking. The result is Not Good.

Re:My take on JDK 1.4

[Glock27]

OK.. I am a Java fan... (recently this has been changing though.)

Sorry to hear that, I hope I can persuade you a bit back towards the good side of the force... ;-)

I have mixed feelings with JDK 1.4.

I'm mostly ecstatic about it. =)

The assert facility is OK.... i don't like the fact that they added an Assert keyword but I don't get to make the decisions.

I wish they had done a better job of support DBC, but it should still be useful. I'm not too sure what it bought over just using existing "assert" hacks, which can also be compiled completely out.

There is also some controversy.

Perhaps in the minds of some...

The Log4J people are upset because there is now a 'stanard' Java package for logging. IMO the 'standard' package is inferior to Log4J in many situations.

I'm not sure what the complaint is here. If Log4J is superior (I've not looked at the logging API yet), it should still attract lots of users. Regardless, there is no way for Sun to extend the core JDK without potentially stepping on the toes of some developers. I don't see an easy way around that problem.

The regexp package is not all it is cracked up to be either. I would recommend Jakarta ORO or Jakarta Regexp.

Again, I can't comment on the technical merits of one versus the other. However, again I'm not sure what your problem is. There are alternative packages, which are better. Great, then use them! The core packages are essentially developed by committee, and are only one approach to the given problem domain. There are many tradeoffs in library design...

As far as that... it runs GREAT on Linux. Probably the most SOLID VM I have ever run.

That sounds great! I'm just about to install it.

They did break some stuff with legacy code. If you ever named a class 'URI' your code will now fail to compile because they put this class in the java.net package which everyone imports anyway.

That is not "breaking something". That is the reason Java has namespaces. Not a big deal at all.

As far as C# vs .Java. I am really impressed with the CLR/CLI stuff.

I'm not sure why...care to clarify? Have you read the comments about CLR based languages all being different skins of C#?

I'm also not sure why people seem to feel it's impossible to interface Java to legacy languages. People do it all the time using JNI.

Right now, as it stands, Java is a proprietary language. Unless we see SUN Open Source Java (or push it through a standards committee), we *may* see a JDK 1.5... but no one will use it.

Sun has made noises about open sourcing Java at some point. If it'd just release the compatibility tests as open source, that would go a long way. (It is currently perfectly fine to 'clean room' Java, as gcj is doing. You just can't call it "Java" without essentially licensing Java and getting the compatibility tests.)

Regardless of Sun's timing in releasing Java as open source, you're completely out of your mind if you think interest in Java is going to significantly lessen because of the introduction of C#/CLR. C# and the CLR are both fledgling technologies, where Java is quite proven and robust. I just read about the first CLR security exploit, and I expect many, many more. Heard of any with JVMs lately?

C# and the CLR only run on small computers and operating systems compared with the high availability big iron that runs Java. Also, Java has penetrated small devices very effectively, with over 100 million Java enabled cell phones to ship this year, very significant embedded penetration, and instruction set support for Java bytecode in all the new generation of ARM chips such as Xscale. Microsoft is caught in the middle (granted the middle covers a lot of ground;).

Despite all the rhetoric to the contrary, I advise you to expect big things from Java on the desktop as well. Sun is quietly staying the course, and is making big improvements to client side Java with every release. Since Java supports Windows, MacOS and Linux, it is a very attractive way to deploy applications (we'll see when C# achieves this milestone;). Earlier releases coupled with slower computers were painful, but things have improved to the point where Java is perfectly usable for lots of applications. Linux + Java is a potential Microsoft killer, make no mistake about it and Microsoft knows it.

My final point for today regarding the "open" nature of C# and the CLR is that quite a lot of the C# class libraries available under MS Visual Studio are not included in the standard. This includes the GUI libraries, so as things stand you can use Java today to write completely cross-platform GUI based programs. It will most likely be a cold day in hell before Winforms are available anywhere but Windows. I'm betting Microsoft's legal team will make sure of it. We'll see if Mono gets far enough to be a test case.

Sigh, now the long wait begins for JDK 1.5. ;-)

(Reptile looks cool, I'll check it out later.)

299,792,458 m/s...not just a good idea, its the law!

Re:My take on JDK 1.4

[Kerg]

So they didn't clone the log4j API, big boo hoo.

Well, they really did clone the log4j API (with a few exceptions here and there), changed all of the names, and added a bunch of log-levels that no one will ever use. You can probably write a sed script to translate between the two packages.

So if they did clone the API, and switching between standard logging and log4j is a matter of running a simple script against my codebase, then what is the problem?

Re:My take on JDK 1.4

[damm0]

As far as C# vs .Java. I am really impressed with the CLR/CLI stuff. Right now, as it stands, Java is a proprietary language. Unless we see SUN Open Source Java (or push it through a standards committee), we *may* see a JDK 1.5... but no one will use it.

I'm not sure how much C# coding you've done. In recent months I've written quite a bit of C# and .NET code. Let me enlighten you with regards to the implication that C# is either not proprietary or cross platform.

It may be true that C# - the language - is cross platform and may acheive recognition as a standard of some sort. However, the .NET framework (which almost every C# program written uses) is not cross platform nor will it be a standard.

Why? Because of small details in the API. Let me give an example. Let's say you want to find out what operating system your program is running on. In Java, you do this:

System.out.println( System.getProperty( "os.name" ) +
" " + System.getProperty( "os.version" ) );

Looks pretty straight forward. In C#:

using System;
...
Console.Writeline( System.Environment.OSVersion.Platform );

Now this all looks very nice. The problem is that the Java version gives you a string, and the C# version gives you a... enum. That's right folks, to find out what operating system you are running, the .NET platform gives you an enumerated integral constant. This means that you have three choices for which operating systems you can discover you are running on, and which operating system the operating system can tell you it is.

Those three choices are:

1. Win32NT
2. Win32S
3. Win32Windows.

(If you want to verify this yourself, check out the System.PlatformID enumeration.)

Now, how do you make this cross platform?

Re:My take on JDK 1.4

[Glock27]

Its a programming language, not a religion.

I don't view Java as a religion - I view it as a technology, which, if adopted across the board, could drastically improve computing (particularly software development, and the quality of software in general). Java also encourages platform independence, and thus solid alternative platforms to Windows. C#/CLR aren't really an alternative, they are simply a (quite transparent) attempt to 'embrace and extend' Java into a component of the overall Microsoft Monopoly.

Therefore, yes, in my world view Java+JVM=GOOD, C#+CLR=BAD. It's really very simple.

You Java people are like a fucking internet death cult.

The true "Internet death cult" (as in "Death to the Internet!") is Microsoft with it's "Windows Everywhere" mentality. If Microsoft succeeds (and .Net is a cornerstone in this strategy), the Internet will become a giant hose spewing money into Microsoft's coffers. I'm sure that will be good for the computer industry! (he says sarcastically)

In my humble opinion, the only people who are pro-Microsoft to any degree at this point are fools or uninformed. Microsoft is anathema.

By the way, it's too bad you couldn't address any of the substantive points I made. Pretty weak.

299,792,458 m/s...not just a good idea, its the law!

Re:My take on JDK 1.4

[Error27]

>>How likely do you think this situation would be if the JCP (or something like it) was not in place?

Perl is open source and it manages to be identical on any UNIX.

>>What I would LOVE is to see Java open sources while ensuring that it remains cross-platform.

Java is cross platform so long as one of those platforms is not Linux. Have you ever tried to distribute a Java program to a Debian user? Java is fine for businesses but for regular Linux users it's not viable.

The whole line that "JAVA must remain proprietary so that it will be cross platform" is the stupidest thing I have ever heard. Name one stituation where being open source made something non-cross platform. Python is cross platform. Perl is cross platform. Glibc is cross platform.

Microsoft has decided that java must die and did not release a JVM with XP. If Sun was smart they would stop spouting crap about how close-source makes things cross platform and realize that they need Linux now more than ever.

Re:My take on JDK 1.4

[scrytch]

> Glibc is cross platform.

*cough* .. erm, glibc requires kernel headers to compile. It works on linux, HURD, and that is it. I'd really love to see it compile and work on win32 or a mac, let alone OpenVMS or S/390

Good articles

[mattscape]

Are there some good articles out there (besides the sun ones) on how to use the new features?
thx
matt

Re:Good articles

[joyrider]

Wrox Press have an Early Adopter J2SE 1.4 book out now, and it's got some good reviews at Amazon...

Re:Good articles

[vanguard]

This article from extremetech is pretty good. The first half is mostly about the swing improvements. Part II (halfway down) covers the rest.

Kodak invented OLE?

[blair1q]

I just took a look at the three patents Kodak is suing Sun over, and, huh?

Sure looks like Kodak claims it invented OLE.

Which, hey, they may have, and Microsoft either licensed or stole it.

What's the real story?

--Blair

Re:Kodak invented OLE?

[briansmith]

Kodac got these patents when it bought Wise. I read on JavaLobby.com's discussion of the lawsuit that Wise sued Microsoft three years ago over these three patents. I don't know the outcome of the alleged lawsuit.

Also, a few people on JavaLobby are of the opinion that Kodac just patented three fundemental object-oriented programming techniques. If that is the case, these patents would never hold up in court as almost any SmallTalk program written before 1990 would be prior art.

Re:Kodak invented OLE?

[blair1q]

I don't think it was just an object-oriented technique that almost any SmallTalk program could map to.

The patents talk about having a UI from one application embedded in the UI for another (like an spreadsheet as a widget in a larger form), and having the two apps communicate rationally about it.

That's enough different from inheriting methods on a class-interface in source code that I can see it as a new patent.

Was there a lot of that going on in SmallTalk in the '80s?

--Blair

My pet peeve.

[Malcontent]

This is in regard to the kodak suit.

What I find especially bothersome is the the fact that Koday (supposedly) had these patents. They did nothing with them. Sun produced a product, hyped it, sold it, improved it and put in millions of dollars and man hours into it.
Kodak then comes in and demands money after the fact when they made no attempt to actually do anything.

I think that's crazy. Why punish the people who got off their asses and did something especially if the punisher was too lazy or stupid to actually make use of their idea.

This isn't just about sun and kodak either. Who was suing palm recently? Same thing.

Sit on your ass doing nothing, wait for somebody else to do all the work. Then sue them and retire in the bahamas. It's the american way I guess. Sure beats working.

Re:My pet peeve.

[AVee]

Read the ZDNET article:
"We've attempted to resolve this with Sun for about three years, but the discussions with Sun have not led to a suitable licensing agreement," Kodak spokesman Anthony Sanzio said.

It looks more like Kodak is suing them because they couldn't work it out in another way.
Whether the patents are valid or not is a different issue...

Re:My pet peeve.

[Kerg]

Considering that probably most object oriented frameworks use something that can be characterized as 'object managers' then I bet Bill Gates *does* do it already.

Re:My pet peeve.

[Bazzargh]

Now that is a crap argument.

1) You have no idea whether Kodak actually use this - the first patent, for example, would seem to cover introspection and/or the java activation framework. It seems likely that a kodak image browser might well use these ideas - just not as publicly as sun.

2) Supposed you were a lone inventor, and patented your idea for eg everlasting bubble gum. Now some big company, say, Slugworth Chocolates, comes along, starts producing your stuff without your permission, promotes it more than you can possibly afford, and thus you never make any money out of your invention.

This is precisely why we have patents. There are many reasons to criticise the patent process, but complaining that a big company is doing more with an idea than its inventor, and thus should be exempt from the law, is not among them.

Re:My pet peeve.

[Jagasian]

Actually, some people like strong steady progress and innovation. Patents slow down the progress and restrict innovation because people effectively own ideas. An analogy to farming would be that many people in my area are starving, and I am the only owner of fertile land. However, I refuse to do anything with the land, I own it, and I want to sit on it.

Patents should be restricted so that they are used only to recuperate money spent on R&D of the idea/concept/technology. This way researchers could, in theory, fund their own research, once they got the ball rolling. Under these restricted patents, your patent last for a very small amount of time, maybe a few years, and it can be cut short if you make enough money off of the patent so as to recover your R&D costs.

Re:My pet peeve.

[alcmena]

That doesn't make much sense. You have to recoup more than your R&D costs so that you can support failed projects. Many projects fail and those costs have to be passed onto the successful ones.

Re:My pet peeve.

[Captn Pepe]

You shouldn't dump too much on patent owners who aren't fully utilizing their patents -- most small inventors (whether they patent their invention or not) see large corporations swoop in and seize their work before they could possibily have capitalized on it. If they're lucky, they have a patent and a good understanding of patent law, and can thus make something off of the corporation in question. Usually not, though.

You probably don't want to strengthen patents any more than they already are, because we're already seeing all kinds of problems with software patents being used to lock open source solutions out of various areas. Many industries also have the problem that start-ups are impossible because the established players already own the necessary patents, and have no interest in licensing them to a new competitor.

Personally, I suspect that the answer is probably a compulsory license regime for patents. In this case, a sensible solution might be to set default payments which are somewhat high, but that scale with number of units sold and price charged. Thus, large corporations still have an incentive to negotiate with patent holders for lower license fees, but start-ups needn't pay anything until they start shipping units, and would be free to use the patents at that rate whether or not their competitors want to let them into the market. Finally, free software would be protected, since in this scheme the developers would be implicitly accepting the default terms, which wouldn't require payment for copies not being charged for (but RH et al would have to fork out for distros that they sell).

Unsurprisingly, a number of powerful lobbies have ensured that this cannot happen without major changes to our IP system; for one thing, it would require breaking the WIPO treaty, which the megacorps paid really good money for.

Re:Not meaning to troll but....

[knulleke]

If I had mod points, I would give them all to you.

Yes java is the most overrated language next to esperanto only.

Security export rules

[Aceticon]

As usual the "import control restrictions" once again are in full force. From the release notes:

Due to import control restrictions, the JCE jurisdiction policy files shipped with the Java 2 SDK, v 1.4 allow "strong" but limited cryptography to be used. An "unlimited" version of these files indicating no restrictions on cryptographic strengths is available.

The JSSE implementation provided in this release includes strong cipher suites. However, due to U.S. export control restrictions, this release does not allow alternate "pluggable" SSL/TLS implementations to be used

What seems even stranger now is that you cannot used "pluggable" implementations (in JSSE). Maybe the "nasty" europeans/asians/africans/martians would provide some unlimited strong cryptography pluggin otherwise???

Re:Security export rules

[JohnA]

There is already a clean room, open source (BSD License) implementation of the JCE. It's called Cryptix, and simply put is one of the best libraries ever written for Java.

I don't trust black box cryptography... especially when Sun goes the extra mile to obfuscate their default implementation of the JCE crypto modules.

Re:Not meaning to troll but....

[mccalli]

What good is this platform really?...It seems to me that Java is nothing but slower than other languages

The platform is not the language.

Java is good partly because of its pragmatic syntax (C++ish with some sugar added, some sugar taken away), but mostly it's good because of its excellent class library.

Though I haven't written anything serious for a year or so due to a job switch, I used to write large-scale multithreaded network servers, where somthing like three to four hundred threads could be running at any given moment inside the server. Java's class library made this really quite easy, and it's syntax is pleasant enough to work with.

Cheers,
Ian

Patent titles

[jeti]

The article says nothing about the nature of
the supposedly infringed patents. Here's their
titles:

US05206951
Integration of data between typed objects by mutual, direct invocation between object managers corresponding to object types

US05421012
Multitasking computer system for integrating the operation of different application programs which manipulate data objects of different types

US05226161
Integration of data between typed data structures by mutual direct invocation between data managers corresponding to data types

Re:Patent titles

[wackysootroom]

US05421012
Multitasking computer system for integrating the operation of different application programs which manipulate data objects of different types

Hmm... I guess that since I Create C++ Programs that run on a Linux and Windows OS, I am in violation of this copyright.

On a side note, if Kodak wins this, they will probably go after MS next due to the OOP syntax and the CLR of .NET

Re:Patent titles

[Corrado]

Wow, can they get any more vauge than that!?!?

Re:Patent titles

[KidSock]

US05206951
Integration of data between typed objects by mutual, direct invocation between object managers corresponding to object types

S05226161
Integration of data between typed data structures by mutual direct invocation between data managers corresponding to data types

These should be an example of why software patents are ridiculous. These two only differ by the changing objects to data structures.

Re:Yah, will this be stable on Linux?

[satanami69]

Okay, okay, although it may have crashed, it's still secure. Probably more so now!

Java Secure Socket Extension

[srichman]

The Java Secure Socket Extension has provided TLS/SSL support to Java for a long time, and is now part of 1.4.

???

Re:Java Secure Socket Extension

[Anonymous Coward]

JSSE is integrated, and does suport SSL, but the point being made is that fact that JSSE was not updated to support non-blocking I/O. So all the talk about how non-blocking I/O is gonna be this great boon for server programmers is somewhat moot, since they don't do SSL outta the box and developing SSL isn't as easy as just encrypting data.

There is an RFE I entered about this a while back http://developer.java.sun.com/developer/bugParade/ bugs/4495742.html . If you're a Java developer, register and vote for it, to make sure Sun corrects this oversight.

Re:Java Secure Socket Extension

[btempleton]

Sorry if this was not clear in the original post, but I was talking about doing SSL over the new "channels" I/O package just introduced in 1.4.

Java has been poor for servers in the past because you had to have one thread per open socket, and the JVMs did not handle very large numbers of threads well.

The answer was to support "select" for I/O on multiple channels without a a lot of thread overhead, but they did not support security with it.

Re:Yah, will this be stable on Linux?

[Wraithlyn]

No shit... 1.3.1's memory management and garbage collector are whacked.

new in 1.4: public Exception(Throwable cause)

[_am99_]

My favorite thing about using 1.4 is code like this:

public void methodA throws MyException {
try { Driver d = Class.forName(driverClass).newInstance(); }
catch (Exception ex) { throw new RuntimeException("problem loading driver"); }
}

can now be this:

public void methodA throws MyException {
try { Driver d = Class.forName(driverClass).newInstance(); }
catch (Exception ex) { throw new RuntimeException("problem loading driver", ex); }
}

Notice the RuntimeException constructor now has the original exception passed to it. It can be retreived higher up the stack, and I believe is printed during a ex.printStackTrace(...). It lets you pass the root cause exception up the stack trace, while preserving the entire state, without having to declare it everywhere.

Re:Open Source Java VM & class libraries

[mlinksva]

Look at the bottom of the classpath page for a list of open source VMs that work with classpath (a free core (i.e., java.* and a few others) class library). All works in progress. I expect that mono and/or portable.net will quickly outpace free java projects. The JDK is a case where the availability of good enough gratis software has seriously hampered the momentum of libre competition. Netscape 4.x is another such case.

Ruby vs. Java

[srichman]

I'm a big fan of Ruby, and a diehard Java devotee.

But I can't forsee Ruby supplanting Java for large projects. The typing is too dynamic, and this ends up being a big headache and source of problems in larger code bases. More concertely, the lack of compile-time type checking makes it hard for Ruby to scale to big projects. You don't find out until runtime if something is type correct, and even then maybe not until some rare execution sequence occurs. Or, worse, it might be type correct in the Ruby sense (i.e., an object can receive a certain message), but not be at all type correct from the programmer's point of view, which might manifest itself in difficult-to-find bugs.

This is a problem with dynamic casting in Java/C++, too, but in those languages the dynamic type checking is the exception rather than the rule (this will get a lot better when Java introduces parametric types in 1.5). More fundamentally, though, at least those languages offer compile-time type checking support, whereas Ruby does not and cannot (since code can be dynamically injected into objects).

Anyone here read the license?

[asb]

5. Notice of Automatic Software Updates from Sun. You acknowledge that the Software may automatically download, install, and execute applets, applications, software extensions, and updated versions of the Software from Sun ("Software Updates"), which may require you to accept updated terms and conditions for installation. If additional terms and conditions are not presented on installation, the Software Updates will be considered part of the Software and subject to the terms and conditions of the Agreement.

6. Notice of Automatic Downloads. You acknowledge that, by your use of the Software and/or by requesting services that require use of the Software, the Software may automatically download, install, and execute software applications from sources other than Sun ("Other Software"). Sun makes no representations of a relationship of any kind to licensors of Other Software. TO THE EXTENT NOT PROHIBITED BY LAW, IN NO EVENT WILL SUN OR ITS LICENSORS BE LIABLE FOR ANY LOST REVENUE, PROFIT OR DATA, OR FOR SPECIAL, INDIRECT, CONSEQUENTIAL, INCIDENTAL OR PUNITIVE DAMAGES, HOWEVER CAUSED REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF OR RELATED TO THE USE OF OR INABILITY TO USE OTHER SOFTWARE, EVEN IF SUN HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

Re:Anyone here read the license?

[TheAJofOZ]

Notice of Automatic Software Updates from Sun

This is a feature of Java WebStart - application developers can specify a minimum JRE to use and WebStart will automatically download it when it is not already present. Note that while Web Start has just become standard with JRE1.4, it will run even on JRE 1.1 and I would imagine this clause is in the separate Java Web Start download as well.

Ironic/NBIO

[srichman]

I used to write large-scale multithreaded network servers, where somthing like three to four hundred threads could be running at any given moment inside the server. Java's class library made this really quite easy, and it's syntax is pleasant enough to work with.

It's kinda ironic that you should say this, since threads are the wrong way to write "large-scale" network servers, and since Java 1.4 finally gives us non-blocking IO APIs to implement things the right way. (The NBIO APIs in 1.4 are, incidentally, largely a product of the work of the fellow behind the second link I gave.)

Re:Ironic/NBIO

[mccalli]

It's kinda ironic that you should say this, since threads are the wrong [usenix.org] way [berkeley.edu] to write "large-scale" network servers

Read the post again. Didn't say I used threads to handle network IO.

By the nature of the app, we had a single thread handling all network IO - it was a single feed of financial data. This data was built up into an internal queue to get it off the network as quickly as possible. Following that, a subscription thread would post notifiers to all objects that had subscribed to this data that information was available, and it would take it off the queue and hand it to the object. Each object was its own thread, and what operations it did were up to it (usually calculations followed by writing to a database).

Non-blocking IO would be a non-issue here - the network is not the bottleneck, and it isn't there that the hundreds of threads operated.

Cheers,
Ian

Re:Ironic/NBIO

[srichman]

Read the post again. Didn't say I used threads to handle network IO. ... Non-blocking IO would be a non-issue here - the network is not the bottleneck, and it isn't there that the hundreds of threads operated.

Read the links. Event driven programming isn't about network performance. It's about the fact that thread context-switching (and, if present, lock contention) overhead subtantially degrades a server's performance under heavy load.

Re:Ironic/NBIO

[srichman]

But they've also found that anything that hits the disk, or hits other slow services (e.g. back-end databases, transaction processing) works better in a multithreaded mode.

I don't think you're understanding it. Things that hit slow services (like databases) do not run better in a multithreaded server; they run better in an event-driven server, the same as for a cached workload. Why on earth would MP/MT be better just because the tasks were slow?

The reason that highly disk-bound workloads do well in a multiprocessed or multithreaded (kernel threads, not user threads) server is actually because very few operating systems provide non-blocking IO for disks. Nonblocking disk IO is more challenging for a kernel to implement, since you have to keep track of buffers and states outside the kernel call stack. For network IO, the buffers are all already there in the kernel, so NBIO is relatively trivial: just let the application know when data has arrived. Therefore, you can't write a pure event-driven server that does lots of disk IO under most OSes, because your server will block (which violates the prime directive of event-driven programming). That is why Flash spawns helper processes to handle disk IO.

If you're working with "back-end databases or transaction processing" (which occurs over the network), then NBIO is no problem, and you should write a pure event-driven server. If you're writing a server that involves the disk, then you should use a hybrid scheme like what is proposed in SEDA.

The main point that we get from these papers (particularly the more recent SEDA), though, is that, if you have to use threads, your hybrid design should use a thread pool of a fairly small number of threads, and you should use events and event queues for the rest. The poster I was replying to said that "three to four hundred threads could be running at any given moment inside the server." I don't need to know anything about the particulars of his application to know that he's going to lose with his "large-scale" implementation: look at figure 2 in the SEDA paper from SOSP. Notice how the throughput starts to seriously degrade when you get past 8 threads?! And how, by the time you get to 128 threads, the throughput is as bad as when you have 1 thread! What does the graph say when we reach the 300-400 threads that our subject's server had? Notice how the throughput has dropped to well below half the single-thread throughput, and the latency is a factor of five greater than the linear latency? This ungraceful service degradation under load is inherent to thready servers, and is largely independent of whether your threads are blocking for the disk or twiddling their thumbs--there is no reason to ever have this many threads. The only point of multiple kernel threads in disk-bound workloads is to give the disk enough work to do to make its read batching and arm scheduling efficient. For this purpose, you don't get any additional returns for having more than about a dozen outstanding reads at any time. Notice how SEDA's webserver caps the number threads in the thread pool of each of its three stages at 20, and the thread pool controller keeps the optimal size far below this on average.

Please read your own citations and call out the domain of applicability of such comments.

Please understand the content of the citations before presuming to school me on them.

So please don't slam someone for having used an approach that is superior in many domains

So, in conclusion, having 400 threads in your server isn't superior in any domain. And I wasn't "slamming" him at all: 1) most servers in the world are multithreaded, so I hardly have reason to be particularly critical of him, and 2) Java didn't even have NBIO when he wrote the server a year ago, so the event-driven design wasn't an option anyway (although you should probably have a much lower cap on the size of your thread pool; Apache won't spawn endless processes until it kills your machine).

Re:Ironic/NBIO

[srichman]

Only when they argue with me.

Re:hope swing is faster now / I prefer Ruby over J

[raistlinthegreat]

I have read the articles. There have also been a lot of these articles when jsdk 1.3 came out and Swing was not much faster. Swing is a cool idea. If it could match the performance of VB on windows or Qt in Linux there would be much more cross platform apps and this would also help linux onto the desktop.

Re:Why is this not under the developers section?

[glwtta]

A wild guess here, but maybe because it is under the Java section?

Re:skeptical for the desktop

[TheAJofOZ]

Sun is like Microsoft - they make as many features as possible, worry about the bugs later and let hardware catch up to overcome their crippling defects caused by overzealous and misinformed OOD.

Okay, I can't take this any more - Java is not slow or buggy and can produce applications that are indistinguishable from C/C++ applications. The proof of this is simple - rewind the clock back a year to MacWorld where WorldBook Encyclopedia was demoed as part of the keynote with everyone watching and they were all impressed. Months later I was informed by the "head Java dude" at Apple on his Australian tour that WorldBook is completely written in Java - but noone knew.

Most people think that Java is slow, buggy and doesn't look platform native because they assume that anything that is fast, stable and looks and feels platform native wasn't written in Java. It's even cooler though, because the "write once, run everywhere" of Java actually works if you write good code). Sure there are some things that Java can't do that you need native code modules to handle but that facility is available.

The worst thing is that young programmers are led to think that Sun's code is actually *good* which spreads their poor, inefficient form to the next generations.

Java's not perfect, but I seriously don't think that it's Java that corrupts young programmers. You can write good code in almost any language - I see an awful lot of really bad Java code written by C programmers but I'm not about to claim that C is corrupting old programmers.

Re:Oil Company

[King Of Chat]

It's a shame that Kodak/Wise never enforced the patents to stop MS from developing OLE. That would've saved many people's sanity.

You agree to all future licence agreements.

[andaru]

"the Software may automatically ... install... ("Software Updates"), which may require you to accept updated terms and conditions for installation."

This means that they are claiming that if you agree to this licence, you are also agreeing to the licence of any application which installs itself (and you have agreed to let it install itself).

I'm sure that that is not legal, and could invalidate the whole agreement. On the other hand, I don't see how a licence agreement that you haven't signed is worth the pixels its printed on.

To say that Sun's online license agreements are binding would be to say that you are not allowed to click on a button with two words on it which is displayed on the web without agreeing to this contract.

As someone pointed out in response to another article, not agreeing to the licence agreement preserves your right to click on the "I Agree" button without agreeing to the licence agreement. It is in the agreement that it says that you agree that clicking the button means that you agree.

Is that convoluted enough?

My pet hates with Java

[divec]

• Unicode support is broken. Java's 'char' type is supposed to represent unicode characters. However it is only two bytes wide so does not support codepoints beyond U+FFFF, such as various Chinese characters. String handling is built around the assumption that UTF-16 has one character per byte-pair, so String.length() can give incorrect answers, String.substring() can give invalid UTF-16, etc. (This fault applies to C# and .NET too, AFAICT)
• Because Java contains a lot of special-casing for String objects (such as overloading '+' and forcing a toString() method in every class), and because String is final, you can't write your own class BetterString which fixes this unicode problem and still works like a String.
• There is no universal type like void* in C. There is Object, but primitive types (int, char, float, etc.) are not objects. You can't pass them by reference. There are duplicate, object wrappers for these (called Integer, Character, etc.) which can be passed by reference, but you spend stupid amounts of time wrapping and unwrapping primitives.
• A function cannot take variable arguments of arbitrary type. The closest you can get is a function which accepts an array of Objects. (But Object is not a universal type - you can't give primitives to such a function without wrapping them).
• You can't create a list anonymously. if x, y and z are ints, and "average" is a function which takes an array of ints, you can't do "average({x,y,z})". This makes the previous limitation more serious.
• Things that should be warnings are in fact errors. For example, this snippet won't compile, because "a might not have been initialized": int a;int b = 0;if(b == 0) {a = 1;} System.out.println(a);. There are times when you want to write code logically equivalent to this, but the compiler makes you pointlessly initialise the variable at the start. The programmer may know things that the compiler doesn't; in this instance, that a will be initialised. It would be better for the compiler to merely warn, and then a runtime error to occur if the programmer was wrong.
• The syntax is intentionally inflexible. To add a worthwhile "assert" statement, Sun had to add it to the core language, which is something that only Sun can do. Compare C, where anyone can write a reasonable assert macro. The same thing will apply to other things which you might want to do, but can't.

Oh yes, and I nearly forgot a trivial-but-annoying one:

• byte is signed in Java - instead of running from 0 to 255, it runs from -128 to 127. If you want to write the byte 0xA9 ( = 169) to a binary file, you have to call it -31.

Re:My pet hates with Java

[_underSCORE]

I fail to see how a declaration "String str = null" is any safer than a declaration "String str".

You've used this serveral times as an example. You're right, assigning a string to null is not safer than leaving it uninitialized. However, that just gets around the 'uninitialized variable' error from the compiler. You should initialize it with something meaningful, or at least "". That would be more safe, and would be what the compiler wants you to do. This error has saved my butt more times than I care to remember, so don't knock it.
If you want to use variable arguments, pass in a vector or a list. There are ways to make java do what you want to do, they might not be like those in C or C++.

Re:My pet hates with Java

[toriver]

I guess it's pointless to argue with trolls, but:

Java's 'char' type is supposed to represent unicode characters. However it is only two bytes wide so does not support codepoints beyond U+FFFF, such as various Chinese characters.

So you didn't know that the Unicode standard quite explicitly states that "Unicode character values are 16 bit"? Hence, no "codepoints" beyond U+FFFF exist!

You can't pass them by reference.

I have seen exactly one method that is dragged out as example of why you need to pass primitives by reference, and that's swap(a,b), which is just of academic interest.

A function cannot take variable arguments of arbitrary type.

Then use Object[]. "Variable arguments of arbitrary type" is a hack that works in C because C isn't properly strongly-typed, and has no place in Java.

You can't create a list anonymously.

As others have pointed out: Yes, you have been able to since JRE 1.1 (prior to that, only in declarations).

The syntax is intentionally inflexible. [...] Compare C, where anyone can write a reasonable assert macro.

Gee, how "standard" that sounds, so now I have to learn some nincompoop's particular fancy if I take over their code. Macros suck anyway because they ignore the type system, and most C++ literature I've seen wisely discourages their use.

As for flexibility, are you opposed to standards? Then stop using C because its syntax is also "inflexibly" specified in a specification.

byte is signed in Java

All numeric values except the special-purpose char are consistently signed. No need to go and check every time you use a numeric variable. This is a good thing.

Unit testing

[Aapje]

I've never used assertions, but they seem a great complement to unit testing. Unit testing allows you to write code to test your functions and easily see if something breaks, the major problem is that they lack an easy way to look inside objects to keep an eye on internal consistency. Assertions can be great to catch those silly little boundary mistakes.

A good unit testing framework for Java is JUnit, they are available for other languages as well.

BTW, you can create your own assertions with Log4J, so even JDK 1.1/1.2/1.3 users can use them:

if(logger.isDebugEnabled())
if (bla>10)
logger.warn("bla>10, bla=" + bla);

This uses almost no CPU-time if debugging is disabled. Log4J is a very good logging package, it surely beats System.out.println, check it out!

Re:Unit testing

[Malc]

How x-platform is that?!

I hope the security model stops unsigned applets from doing that - the NT Event Logs can filled up very quickly.

Re:skeptical for the desktop

[TheAJofOZ]

I'm not trying to criticize Mac users here at all, I'm just saying that impressing people at MacWorld doesn't prove much at all.

Go and look at WorldBook on OS X and you will also be impressed. More importantly though, noone realised it was a Java application. It looked and acted just like a C/C++ application.

Re:Java2 ?

[grantm]

Remember, this is Sun the people who released Solaris 2.0 which:
- was the first version of Solaris
- declared itself to be SunOS 5.0
- was an implementation of System V Release 4.0

Obviously being able to count is not a prerequisite for getting a job in Sun's marketing department.

Re:skeptical for the desktop

[E-prospero]

Most people think that Java is slow, buggy and doesn't look platform native because they assume that anything that is fast, stable and looks and feels platform native wasn't written in Java

I find the example set by WorldBook 2002/MacOSX to be a stunning demonstration of Java's potential if Sun would just wake up and smell the coffee (pun intended).

I like Java a lot - both as a language, and as a bytecode platform. I use it daily at work, and at home for my own projects (after many years of flipping between C, C++, and a dozen other niche languages)

My only major bugbear with the JDK is not with Java itself, but with Swing: It's godawful slow (The Hello world app should not take 30 seconds to display on a P400), it's API is occasionally appalling (JTree anyone?), and it has a habit of being occasionally buggy.

Above all, it reinvents the wheel with widgets. The Look and Feel implementation very occasionally looks and feels like a platform native application (Win2k LnF anyone?). However, it usually looks and feels like someone has tried to reimplement a system native toolkit using only a blunt light blue crayon.

For many a year I have whined to all that would listen "Why the *@*)(!&*@ doesn't the Java widgeting toolkit defer to system native widgets for window display? Surely this would look better and run faster than pixel blitting a widget look-a-like!"

MacOSX provides just such an API. Although Objective-C is the `preferred' language of Cocoa, there are Java bindings. Note - this API is NOT Swing - it is a MacOSX extension API. Consequently, a MacOSX application built in Java should be almost indistinguishable from a native compiled app in terms of look and feel. And according to your comments, performance is also indistinguishable.

I have played with Java bindings for Gnome; they provide blindingly fast gui performance, using the same java runtime as is used by Swing. However, there are several java binding projects for Gnome (all partially complete), and none of them really address the general problem of widgeting across platforms.

I have also played with Eclipse, the IBM Java . Their widget toolkit is cross platform in API but system native in widget; however, I have found the performance of the Eclipse widget set to be almost as bad as Swing. However, it is beta code - we may have to wait and see if anything improves with age.

Java/Gnome demonstrates that it can be done under X. Eclipse demonstrates that it can be done cross platform. MacOSX demonstrates that it can be done well enough as to be indistinguishable from system native widgets.

So can anyone tell me: WHY hasn't it been done? (And don't say its because they don't want to break backward compatibility - Look at 1.4's NIO framework and VolatileImage).

Russ Magee %-)

Headless at last - bye bye XVfb

[Taurine]

My favourite new feature is that lightweight components can now be run in headless mode - see

http://java.sun.com/j2se/1.4/docs/guide/awt/AWTC ha nges.html#headless

You have to set a property to true:

-Djava.awt.headless=true

as a switch when running the VM for example. Then you can generate server-side graphics on Unix without having to run XVfb. This has been an annoyance for some time, as you had to have different deployment rules depending on your target OS, as NT always has a graphical environment taking up resources whether you want it or not, so it wasn't an issue there.

The upshot is that you can now use java.awt.Image.BufferedImage as an image source in servlets that generate dynamic images, instead of java.awt.Frame, which always seemed wrong. It uses less resources too, as it doesn't have to do a context-switch to your X server to create images!
Hey, there are lots of great new features in this new release, but that is the first one that made life easier for me.

Re:Headless at last - bye bye XVfb

[Fjord]

You've always been able to use java.awt.image.BufferedImage to create dynamic images. The application I'm working on does that. But you are right that now you don't have to use xvfb (which I never thought was that big of a pain, but I like the -D solution better).

I'm kind of interested in java.awt.image.VolitileImage. It makes an image using the video card's memory, with the trade off that the image may be destroyed at any moment. I've done some tests and on my machine it didn't improve performance at all, but on a system with a good video card, it may be an awesome class.

Re:skeptical for the desktop

[TheAJofOZ]

MacOSX provides just such an API. Although Objective-C is the `preferred' language of Cocoa, there are Java bindings. Note - this API is NOT Swing - it is a MacOSX extension API.

Yes and no. The Cocoa APIs are an extension that is not cross-platform, however if you use the Swing libraries on OS X, by default they use native widgets. So you can write a 100% pure Java application that uses native widgets on OS X. I would expect that this kind of thing will flow back into Sun's JVM around 1.5 or so.

Yeah, But...

[Greyfox]

You can use templates to make C++ put on a pink tutu and dance a little dance for you. Alexandrescu uses them, for instance, to make C++ forcibly multiply inherit a chain of classes into a single child class. He also has a single inheritance template setup and a lot of other sick and twisted stuff that every C++ programmer should at least know about.

Re:skeptical for the desktop

[E-prospero]

Yes and no. The Cocoa APIs are an extension that is not cross-platform, however if you use the Swing libraries on OS X, by default they use native widgets.

Ok - this was not my understanding. I thought the OS X Swing libraries were an Aqua-esque LnF, but not Aqua widgets themselves. However, I'm not a Mac junkie, so I don't speak from experience on this point. If they are native widgets, it certainly bodes well IMHO.

I would expect that this kind of thing will flow back into Sun's JVM around 1.5 or so.

I certainly hope so...

Russ Magee %-)

Shouldn't be any harder

[yerricde]

10 Let M$ = "Microsoft"

.NET will probably be much harder to code by hand than java, or VS.NET would not be so necessary.

It shouldn't be any harder. You don't need an IDE to write in the Java language for the J2SE platform, but Sun still provides a free(beer) IDE that includes a form designer. Likewise, you don't need an IDE to write in the C# language for the .NET platform, but M$ still provides an expensive IDE that includes a form designer. Besides, the languages are so similar that it may even be possible to take simple business logic (not I/O) written in one language and simply recompile it in the other language.

Why I Don't Want to Touch This

[Greyfox]

I use Java at work and I had to beat 5 fucking team leads with a crow-bar to convince them to move to Java 1.2. Now they all run the other way when they see me coming -- it's a lot harder to get near them with a crow-bar. By the time my company gets done with its final focus meeting to decide that moving to 1.4's a good idea, Sun will be releasing Java 2.4 (With telepathic user interface code.) If I get all used to how great Java 1.4 is, work will suck more, and it already sucks so much that light can't escape.

Re:Problem with Asserts

[DahGhostfacedFiddlah]

Depends on the situation, of course. Exceptions are (relatively) expensive.

Still didn't fix it...

[Xawen]

Ok, so 1.4 may be faster. 1.4 may be more stable on Linux. 1.4 may be more robust. But, they still didn't fix the thing that annoys me the most about Java. Why is it so freaking complicated to do a simple read from the keyboard???

Re:skeptical for the desktop

[TheAJofOZ]

However, the Java built-in libraries have enough built-in bloat to make further promises of speed indistinguishability just plain misleading ("it works fine, just avoid this list of 50 poorly implemented classes"). No, Java the language isn't slow, but Java the set of Sun library implementations has many, many deficiencies.

I would again have to contest this. WorldBook makes very heavy use of graphics (as does the stuff I work on) and it still managed to perform well and didn't seem to have any problems with bugs. There are bugs in almost every library and I have not found there to be any more in the Java APIs despite constantly working with the latest betas and early access releases.

I took some of my company's Java apps that ran well under Windows and tried them under OS X and they were dog slow. At one point I managed to get ProcessViewer to report CPU usage over 100%! Highlight some text in your browser - that's what it does. Two simple math coordinate checks, a couple calls to drawRect and drawChars, written about as efficiently as Java allows but with every set of mouse movements I saw the CPU usage shoot through the roof.

I would be very suspicious that when you wrote your custom class you optimised it based on your knowledge of Windows and not of Mac OS - the two graphics subsystems work very differently so your optimised code is likely to have actually been the worst possible code for OS X. If you want to benefit from Java's cross-platform abilities you actually have to use them. The API is the key to getting cross-platform compatibility because it is reviewed and optimised for each OS it is ported to. Your code doesn't have a hope of knowing as much about the system or being better optimised. I spend all day running the exact same Java application on Mac and Windows (95 and XP) and it runs the same on both at about the same speed because I take advantage of the APIs.

I know JBuilder runs fine on my machine (I suspect some Mac-specific code or vm settings) but Java isn't at all guaranteed to run well across platforms.

JBuilder is not heavily tweaked for OS X, it's just well written Java code. Poke around JBuilder's files and you'll find it doesn't do much more than a simple 'java -classpath blah com.borland.JBuilder' For the record I find Java runs perfectly fine on everything from a P100 right up to the 1.2Ghz Athlon and from a Rev B iMac (300Mhz) up to the 400Mhz G4 laptop I currently use. Perfectly fine in this case means comparable to programs written in C. Yes there's a longer launch time in most cases (again, OS X makes huge inroads into this and 1.5 will likely adopt this technology to reduce loading times and memory requirements significantly).

On Mac, it looks sweet except for text rendering.

I don't know about you, but I quite like the smooth, easy to read look that Java's text gets on OS X with it's antialiasing.

The bugs are actually probably more annoying because they are the let-downs "Oh, look it has this great gui widget" and then ten days later "Oh crap that widget has a horrible bug and I'm completely reliant on Sun to fix it

I really haven't seen that many bugs in the Java APIs that you would consider it worse than other libraries. If you are consistently having trouble with JTree then it may be worth extending the class to fix the bugs where possible (and you'd be amazed at how much you can achieve this way) or develop your own solution (as a last resort). Consider that with C/C++ you don't get *any* graphical abilities for free (all provided by 3rd party libraries which are completely platform specific etc, etc), you really aren't that badly done by if you have to implement a couple of custom widgets in Java.

When it comes down to it, the solution is to write good code. I can be pretty certain (but not 100% sure obviously) that you are not writing good *Java* code because you are so anti Java. You simply have not accepted the languages strong and weak points and so can't take advantage of the strengths. You also seem to come from a background in another language (C/C++ would be likely) and you don't seem to be willing to adapt to the "Java way" because you are hooked on the "C/C++ way" (or whatever previous language you prefer).

I apologise for any offence in the above statements, I can see from your posts that you are neither stupid nor irrational and in fact are quite intelligent with a fairly open mind. However, changing programming languages is really difficult particularly when they seem as similar as Java and C++. You have to realise that Java is only asthetically similar to C++ but is in fact based on a very different grounding principle. C++ looks for speed and efficiency at the cost of safety (buffer overflow anyone?) and platform independance, whereas Java focuses on safety and cross-platform consistency at the expense of speed (yes there is a speed penalty to be taken but in most cases it is an acceptable trade off and great gains are being made). Further to this, C++ is a procedural language which has had object oriented "stuff" added on top whereas Java was designed from the start to be object oriented (with some procedural elements).

Basically, there is nothing stopping you from developing your own API for Java but I'm pretty certain that you can't do a better job than Sun has - if it was so easy and/or the Java API is so bad, why hasn't anyone created an alternative?

Still no GUI fixes?

[fluor2]

I hate to say it, but I've still not seen a single java application with a lot of windows (using the GUI) perform well under my MS Windows. I know this is not Windows' fault, since they also seem to lag under Linux.

My class and I was at Sun Microsystems in Ireland in 2001. They talked a lot about how good Java was. But when I asked one of the top people this simple question "Why isn't staroffice programmed in Java?", they said "We don't know, actually".

I know why. It's because Java just can't compete with faster languages when it comes to larger programs that require real window handlement. (personal opinion ofcourse).

Now on to the "so-called" platform independability. I personally have been asked several times to convert java programs so they can be run on UNIX platforms (like HP-UX) because they just didnt perform well on those platforms. Instead we chose TCL or similar. Which seem to do the job.

Best luck to Java in the future. Hope you programmers can fix these "small" problems :-).

Re:Still no GUI fixes?

[Knight2K]

Actually, the release notes claim that it has fair amount of performance enhancements to the Java2D API including support for 2D hardware acceleration. On UNIX X-Windows calls are used to implement the Swing GUI classes to improve speed. Sounds like we will see some significant performance gains on all platforms, if you believe the documentation. I haven't seen a Swing app run under 1.4, so I'll reserver judgement.

Re:Still no GUI fixes?

[4of12]

"Why isn't staroffice programmed in Java?"

Maybe Sun/Star took a lesson from Corel, who got onto the Java "platform independent GUI" bandwagon too early.

IIRC, an effort to rewrite a new version of Wordperfect in Java failed because the resulting application was deemed too slow.

Perhaps things have improved since then and Java is carrying the more baggage than it deserves about GUI sloth due to the early overhype [this happens with a lot of projects, IMHO].

Are there other examples besides WorldBook on MacOS that provide more contemporary evidence of Java's speed for interactive GUI applications? For that matter, numerically intensive applications in Java have also been plagued in the past by poor performance.

I love Java as a programming language, wish I could use it more, and hate having to live inside the infested jungle of C++ due to decisions made 7 years ago.

Re: enums and Java

[flagstone]

if I defined all the enum lables as finals in Java, and then used them in switch statements/assigns on variables, I don't get the type-checking I do in C. So in C:

enum PersonType aPerson = Desktop;

would be flagged as an error, but in Java the equivalent

int PersonType = DESKTOP;

would be perfectly fine as far as the compiler is concerned.
This could be a problem.

True, which is why you shouldn't do enumerations in Java that way :-). See this article for a description of how to do typesafe enumerations. However, as is pointed out in the article, you still must do more work in Java than in C++ to accomplish this.

Wording was bad

[SuperKendall]

The story should have read "no LICENSED open source J2EE implementations". There are OS J2EE app servers (JBoss), and in fact they are quite good... the problem (as stated in the article) is that it's hella expensive to get the official seal of J2EE. That doesn't mean it doesn't work!

Now if you think .NET is better, you have to ask how good an open .NET server you're going to be able to build without ASP.NET or Windows Forms. Neither of those are part of the current ECMA submissions, though as stated in the .NET article yesterday they are expected to be submitted at some point...

At the moment J2EE has gone through a lot of refinement, and I think makes a pretty good platform for server side development. I think desktop code is still up for grabs by either Jvaa or .NET (at least under windows). It will be interesting to see if .NET app development is nearly as annoying as MFC was (I doubt it will be).

Re:Problem with Asserts

[Malc]

"Actually they are not expensive at all as long as they don't get thrown. If they are not, it just takes a little extra stack management when calling functions/entering try blocks. "

Actually, there is a cost to exceptions, EVEN if they don't get thrown. This is why MSFT allow the extended syntax keyword nothrow. It's not just "a little extra stack management" as you put it.

The "facade of neutrality" is...

[freeBill]

...what MS employees who get paid to make fake posts on Linux geek discussion groups as anonymous cowards try to erect to cover their tracks.

Two other corrections

[SuperKendall]

Others have corrected some other points pretty well, but I thought I'd throw in two other points I didn't see well covered:

1) Warnings/errors. After many years of real world programming, I'm going to say right now that that is an error. If you find a variable that is not initialized, it is going to be a bug - even if not now, nine times out of ten it will be at some point in the future.

I personally am happy to have that not compile instead of relying on someone else to run lint (or the equivilent) frequently over the code... things like this are one of the reasons Java programmers don't need to run Lint style checks much. The langauge and general programming conventions help elimiate "dumb as a hammer" errors that we ALL make from time to time.

2) Java has some nice internationalization support if you really care to do it correctly for all cases - you can use a BreakIterator to properly navigate through a string for any language.

Java has always had (I think) well thought out tradeoffs between performance and ease of programming that led to things like this, but generally always supported some way to do things "more correctly" if you really needed to - wrapper objects are an example of this is so is String not handling characters wandering past 16 bits in size. You have the primitive type wrappers if you need them (as you noted), and also the heavy-duty internationalized support classes like BreakIterator and Collator.

Re:Open Source J2EE

[lal]

Here's a quote from a Sun marketing person in charge of J2EE branding:

...having a strong brand and compatibility standards are important to the development of a robust market for J2EE platform products, tools, and components. The J2EE Compatible brand has achieved significant momentum over the past two years, and we want to make sure that any open source efforts don't impact the viability of that effort.

My translation: The J2EE brand is like the little Windows flag on the Windows box -- you gotta pay the man to get it.

assert keyword was the right decision

[melquiades]

i don't like the fact that they added an Assert keyword but I don't get to make the decisions.

Dude, read up. There is an excellent discussion of the reasoning behind their design decisions. They give a very good justification for adding a keyword; in fact, it's one of the FAQ questions:

2. Why does this facility justify a language change, as opposed to a library solution?

We recognize that a language change is a serious effort, not to be undertaken lightly. The library approach was considered. It was, however, deemed essential that the runtime cost of assertions be negligible if they are disabled. In order to achieve this with a library, the programmer is forced to hard-code each assertion as an if statement. Many programmers would not do this. Either they would omit the if statement and performance would suffer, or they would ignore the facility entirely.

Unlike a lot of other languages (C++ and C# spring to mind), the Java designers have been very tight about letting multitudes of constructs into the language. Java's minimality and internal consistency is one of the reasons it's been so successful, and its designers know this. They are very intelligent people who are making decisions very carefully -- and they're not going to add something to the language unless they have a very good reason.

In this case, they have a very good reason.

there are never namespace clashes

[mikemulvaney]

Namespace problems are compile-time errors in java. The import statements are just syntatic sugar; it makes it easier for a developer because he doesn't have to type in the fully qualified class names every time. When the source is compiled to byte code, the classes are fully qualified.

For example, if you have code like this:

import java.sql.*;
import java.util.*;

// start up a class, do some stuff, and then...
Date date = new Date()

The compiler will come back with a message like: "Can't figure out if you mean java.sql.Date or java.util.Date". So its not ever a namespace clash, because the compiler catches those problems before they get started.

As far as poor documentation of dependencies, that isn't really a factor either. You can't look at the import section to determine class dependencies anyway, because you don't have to import a class to use it. You might want to make that a convention, but its one of those things that is hard to maintain over a lot of code. If you want to call that 'source-code-level documentation' you can, but that's not what it is. Its not 'source-code-level' any more than the javadoc comments are.

-Mike

Re:Yah, will this be stable on Linux?

[Dr_LHA]

kernel: kernel BUG at pktcdvd.c:1046!

You should only complain about linux kernel being unstable when you run a vanilla 2.4 kernel. pktcdvd.c is not part of the standard kernel, it's part of Jens Axboe's packetised cdr writing patch - which is alpha quality. That patch is not even in 2.5.5-pre1 so you can't really blame linux - only the buggy patched version you have running!

Re:Are you kidding me?

[JediTrainer]

I don't think that's what he meant.

From the console, it would be extremely nice if there was a way that one could read a SINGLE character, without the user having to press Enter/Return.

I've run across this issue ever since Java 1.0, and so far there doesn't seem to be a way to do this. Perhaps there never will.

Re:Open Source Java VM & class libraries

[mlinksva]

I said it is a work in progress. It partially implements APIs introduced in every major Java release (including 1.4) but doesn't fully implement all APIs for any release (perhaps not even 1.0). Check the status of package implementations on the status page or cvs. It's a free software project, people work on what they want to work on. Actually it has picked up some momentum recently with contributions from Red Hat, which is merging gcj library work with classpath.

Re:Yah, will this be stable on Linux?

[RMSIsAnIdiot]

So... to counter that, an NT BSOD is almost _always_ due to a shotty 3rd party driver or strange hardware issue. So we can't really blame windows... only the buggy drivers we have running! So you should only complain about Windows being unstable when it is running a vanilla install.

J2EE Certification

[ahde]

I'm a big fan of JBOSS, and (though I haven't tried it) would've liked to see Enhydra succeed as open source. But I don't see why Sun should grant a J2EE trademark to them. Sure, the price might have been outrageous, but companies like IBM and and BEA seem willing to pay it. Sun allowed them to see the source from the reference implementation and allows them to use the parts of it that haven't been implemented independently yet.

The spirit of open source isn't about marketing. And the J2EE mark is just that. JBOSS is very successful on its own right, and whether Sun blesses it shouldn't matter. Apache has succeeded without Netscape or Microsoft's compatibility certification, though there are still companies who feel more comfortable with a web server branded by a large corporation. Though this draws comparison with Versign refusing to recognize certificates from Apache, its not quite as bad, and eventually, they were forced to accept. Anyway, verisign isn't exactly to be held up as a model corporate citizen

It might be desirable for Sun to open up java to public standardization, but I don't think anyone build on java expecting that to happen. Its like WINE expecting Microsoft to certify it as 100% Windows compatible (even if it really were.)

It may be one day that people will want their J2EE implementations to be JBoss compatible.

Re:Yah, will this be stable on Linux?

[ahde]

Those Aiee!!! messages are what first got me interested in Linux. My dad had an ISP and when I'd go out to the server room (a shed in our back yard) and reset modems, I remember seeing that and just laughing. That was a computer crash that had some flair. Calling the sysadmin and explaining that was fun.

Practical Experience from the Field

[Courageous]

Two thumb up on JDK1.4. The new io (nio) alone is worth the switch. I've been playing with it since beta one, and except for a strange bug (now fixed) in UDP where all incoming packets on all channels each falsely reported their origination address as having come from the same address as was reprorted on the first packet... except for that!:)... this stuff rocks. Packet transmission speeds are up something like 40-50% or more if I recall correctly. It's quick.

The Channel abstraction for talking to sockets also happens to be a nice one. The whole nio library has the taste of having been well-thought out by someone who really new what he was doing.

C//

Re:Problem with Asserts

[renoX]

Actually it depends how the exceptions are implemented.
I remember an article from Borland where they said that one of the difficulties of the porting of Kilyx from Windows to Linux is the exceptions.

On Windows, they are "easy to use" for the compiler's developer because they are a part of the OS but they have a cost even if they are not thowed on Linux they're difficult to use but they have no cost when they are not thrown.

Re:Problem with Asserts

[coyote-san]

You've missed the point of assertions.

The point isn't to catch runtime errors, it's to catch improper usage or coding errors. If your function should never be passed a NULL pointer, you don't want to raise an exception that the caller may catch, you want to bring the program to an immediate halt so you can figure out why the caller passed a NULL pointer. Didn't it check for that? Did it somehow get bad data?

(This use is why many of us define our own assert macros in C/C++. The standard assert() calls exit(), but that doesn't give you any information about the state of the calling procedure from within a debugger. An attempt to divide by zero, on the other hand....)

Another standard use of assertions is to check for coding errors. This is especially useful with loops - if you can identify a loop invariant, you can specify it in an assertion and catch most coding errors immediately.

A related use is to identify class invariants, and to always check the class invariants before you leave any class method. You can quickly identify any code that creates broken objects.

If you're also doing unit tests, you shouldn't need your loop and class invariant checks in production code. However you'll always need to do runtime sanity checks on your parameters, with well defined behavior for what the procedure will do when passed null pointers, negative values, etc.

It's just you

[autopr0n]

I mention autopr0n in my sig, but the majority of my posts here do not talk about it. If you want to see for your self, just go here and read the last 25 commenst I've made.