Gosling Returns To The Java Fold

MemRaven writes "In an article on CNet News, James Gosling reveals that he's returning to the Java Tools group at Sun. The article touches briefly on the Eclipse situation as well as some vague statements about what he's doing in the future. Since he's been gone from the Java fold for a while, this might spell some definite changes in how Sun treats its stepchild."

Comments

great

[n.o.d.y.n.e]

Whats good for the gosling is good for the gander.

Why is this news?

[ajagci]

Let's see what Gosling has done:

He created a commercial Emacs clone, which didn't particularly ingratiate him to the open source community.

He created a commercial window system called NeWS and tried to kill X11 with it, but that was a commercial failure, never really worked anyway, and was largely based on other people's technology.

Then he built a simplistic language for programming consumer devices, but that project failed miserably as well. Only when they put it out on the Internet for free, claimed that they were going to make it "open", and promised to create a browser based application delivery platform did it take off--not because there was anything technical novel about it, but because people wanted to believe in browser-based programming (sadly, Sun has pretty much failed to deliver on all of that). Most of the hard work to make Java a success was done by the JIT developers and IBM.

These days, he seems to be porting over code highlighting and some other features from Emacs to NetBeans.

Sorry, but if this is a "personality story", maybe someone can explain to me why I should be excited about it. At Sun, Guy Steele would be my vote for one of the most competent people they have. But Gosling? Why?

Re:Why is this news?

[ignatzMouse]

It's not how you handle your victories that make you a success. It's how you handle your failures.

Re:Why is this news?

[axxackall]

He is a part of the marketing machine, not of real technologies. Well, this world is driven by over-abused marketing.

Re:Why is this news?

[Anonymous Coward]

"Over-abused marketing"? Is there such a thing as "under-abused" marketing? Or even "abused just the right amount"? Any abuse is, by definition, too much abuse, isn't it??

Re:Why is this news?

[Anonymous Coward]

You think Sun has failed to deliver a decent Java.

Hi, I'm Earth, have we met?

Lots of great software, in *all* categories is written in Java. It's all cross-platform. It all works for everybody. Java is a beautiful version of the type of system that we've all wanted. No need to compile for each computer.

You sir, are too hard to please.

Re:Why is this news?

[Anonymous Coward]

You think Sun has failed to deliver a decent Java.

I said that Sun failed to make browser-based application delivery a success and they have failed to create an open platform in any sense of the word "open" that was in use until Sun tried to change the meaning with the JCP. Whether their platform is "decent", technically or otherwise, is a different question.

Java is a beautiful version of the type of system that we've all wanted.

Java's type system is nothing new and it has some serious bugs (with covariant arrays and parameterized types). I don't think many people have been holding their breath for that.

Hi, I'm Earth, have we met?

Hi, I'm on Earth, too. Do you understand written English?

Re:Why is this news?

[Anonymous Coward]

Good point. In 2004, we should probably update that to "It's how your PR department handles your failures." :-)

Re:Why is this news?

[arethuza]

Well, I have to say, NeWS was a thing of beauty - especially with the HyperNeWS stuff that Arthur Van Hoff did. And Java, like it or not, and I do, was a great tool. If the current explosion of APIs is a bit confusing I really don't think it has much to do with his original vision or, indeed, code. He did good stuff, the man deserves respect.

Re:Why is this news?

[Anonymous Coward]

He said "type of sytem", not "type system". I read it the latter way first, but then I reread it because I couldn't believe anyone would refer to it as beautiful.

Re:Why is this news?

[ajagci]

Well, I have to say, NeWS was a thing of beauty - especially with the HyperNeWS stuff that Arthur Van Hoff did.

I have to disagree. There were several window systems that put code display-side. NeWS perhaps got the most exposure at the time, but it was, in my experience, the flakiest and hardest to program. No doubt, a lot of that was due to the exceptionally poor choice of PostScript as the displaly-side programming language. Even Apple finally exorcised that demon.

And Java, like it or not, and I do, was a great tool.

Java was (and is) 1970's technology. The original Java design and implementation, a typed stack-oriented byte code, was unoriginal and contained some serious blunders. Credit for Java's success goes to Sun marketing who, whether you like it or not, really did do a spectacular job, the Self team and IBM for bringing the technology up-to-speed, and lots of other contributors for fixing innumerable problems with the original design.

It's good that Java has finally brought garbage collection and runtime safety into the mainstream. But that it was Java, rather than one of dozens of similar languages available at the time, is a pure accident of history.

Re:Why is this news?

[arethuza]

Well, I have to admit it wasn't particularly practical. But I did find that once you got the hang of it that interactive graphical programming in PostScript was immense fun. As for the 'originality' aspect - its not something that I care about too much as it was never an academic excercise (and I say that as a ex-academic!). Success in the marketplace is IMHO about having 'good enough' technology at the right time with the right spin. Its not fair, but whoever said it was. You are right - I still think Common Lisp was the 'best' development environment that I have used: CLOS still makes Java/.NET look unfeasibly primitive, but I'd *never* try and persuade someone to do a commercial project on that platform, even if it was technically the right thing.

Re:Why is this news?

[LarryRiedel]

Gosling? Why?

I think it is because of acclamation from sun!wnj.

Larry

eclipse has passed critical mass

[Anonymous Coward]

there's nothing Sun can do about it. Gosling's comments about a developer perspective says it all. This is why eclipse has gained such a great following. I know from first hand, a large percentage of jakarta developers are strickly using eclipse now. If you ask around, the Tomcat developers for the most part use eclipse. Many other jakarta projects use eclipse exclusively.

Eclipse is a slow-moving truck

[schmaltz]

Designed to have massive loads heaped upon it... and move slowly. Am I the only one who remembers when you could run a relatively complex IDE on a 300mhz box without needing to shut down Mozilla or your mp3 player?? Or even a 100mhz machine? What is it with Eclipse, does blinking the cursor take all that much power? Sheesh.

Re:Eclipse is a slow-moving truck

[mohaine]

Java IDEs have always taken a butt load of resources. Some of the really only old IDEs without code completion didn't take as much memory, but what is the point of an IDE without code completion?. May as well use vi.

As for Eclipse, it is pretty speedy as long you your have the ram. 512 is really the minimum on a Win2k + Eclipse box. (200M for Win2k + 100M for eclipse + extra for everything else). From what I've seen Eclipse isn't overly ram hungry for a moderen IDE. It seems to use slightly less memory then IntelliJ Idea, which is Currently at 150M on my box, even though it says it is only using 95M.

Re:Eclipse is a slow-moving truck

[axxackall]

You should never use Java if you don't have enough resources for it. Java was specially designed for embedded devices with several GHz CPUs and few GB of RAM.

For a regular old 300 MHz PC with 128MB of RAM I can recommend either Python/Tkinter or Tcl/Tk for a language and Emacs for IDE.

As for 100MHz PC with 32MB - I doubt that Mozilla can make on it.

Re:Eclipse is a slow-moving truck

[The Cydonian]

IDEa is apparently the leanest Java IDE you can get. Eclipse is actually more lean than most of its competitors (jBuilder, Sun's offerings etc), but yeah, it might create problems if you don't have too much RAM.

Re:Eclipse is a slow-moving truck

[Mr. Piddle]

Java was specially designed for embedded devices with several GHz CPUs and few GB of RAM.

The moderators must have slipped, accidentally selecting "Insightful" instead of "Troll."

Re:Eclipse is a slow-moving truck

[Anonymous Coward]

Java as a language was originally designed for embedded devices. It is not a troll, it's a very well known fact that you can find even in Sun online source.

Then it's failed on that market. This is also not a troll, it's very well known fact you can find on many market reports.

Then it's been moved to Internet. Is it a troll?

At the Internet industry Java was discovered as useless without SWING at GUI and EJB at the server side. It's obviously not a troll. Otherwise corporations would still use AWT and plain (CGI-like) servlets.

And THAT overbloated addons made Java to require GHz CPUs and GBs of RAM. How this can be a troll to anyone who really tried to squizeany performance from either SWING or EJB?

Excuse me, but your comment, without any facts or logical points, is a troll, not mine.

Re:Eclipse is a slow-moving truck

[Jerf]

I've compiled Mozilla ~1.1 on a Pentium 133MHz with 40MB ram in about a day.

It's a close thing, though; LyX failed to compile on that machine because the compiler even ran out of swap. (LyX uses a lot of C++.)

Re:Eclipse is a slow-moving truck

[Anonymous Coward]

Before I upgraded my laptop, I use to use an old laptop with 256Mb of ram. Have you even tried eclipse for more than a second? Hands down, eclipse is one of the fastest java IDE's. I'm still using eclipse on my old 450mhz P3 system and it is responsive. Of course the system has 512mb of ram. Java IDE's are memory hungry. People should know better to confuse disk cache related slowness with application slowness. I actually upgraded my laptop because installing .NET and IIS made my system crawl. Those who think the latest VS.NET is great are right and wrong. It's great if you have 512mb of ram or more. If you don't it's worse than eclipse, because it loads a ton of libraries and eats up a huge amount of memory. In both cases, the disk caching is what kills performance. It's not that VS.NET or eclipse is slow. They're just memory hogs.

Re:Eclipse is a slow-moving truck

[Anonymous Coward]

Visual Studio is quite fast if you just use the command line components. Run vcvars32.bat to set the environment variables. Then use cl and link coupled with Vim for your text editor.

Re:Eclipse is a slow-moving truck

[Mr. Piddle]

And THAT overbloated addons made Java to require GHz CPUs and GBs of RAM. How this can be a troll to anyone who really tried to squizeany performance from either SWING or EJB?

Because they really don't require GHz CPUs and GBs of RAM. They never did. Hundreds of MB for big Swing apps, perhaps, but not GBs. Performance issues with EJB stem from the developers misapplying EJBs and writing code on a pre-schooler level. EJBs simply are not appropriate for small projects, either, which burns many developers still foaming at the mouth with the hype. It was true when Java started and is still true today: bad programmers write bad programs, whether in C or Java or Python or Ruby or whatever. Java itself is hardly the cause of this problem.

Re:Eclipse is a slow-moving truck

[Anonymous Coward]

that's funny. VS.NET is quite fast assuming you have 512mb of ram. 256mb of ram is slow. In fact, it was so slow, I had to make a big stink about about the machine a team member was using. they finally gave him a new P4 2.4ghz with 512mb of RAM. Before that he was using a 1.6ghz system with 256mb. Debugging was especially painful and took minutes for it to start the debugging process. Once he got the new machine, it was sweet, and everything was fast.

Re:Eclipse is a slow-moving truck

[anomalous cohort]

My own findings have been that VS.NET and Eclipse perform about the same. They both are snappy for a while. They they go off on some tangent and the GUI doesn't respond. If you wait long enough, all your keystrokes and mouse clicks will catch up and the GUI is responsive again, for a while.

64-bit Array Indices?

[mosel-saar-ruwer]

Will someone PLEASE tell this genius to add 64-bit array indices to the language?

for(long i = 0; i < whatever; i++)
{
a[i] = foo.bar(i);
}

What in the world good is this stupid language on a 64-bit platform?

And yes, WE DO NEED 64 BITS. LIKE YESTERDAY.

Our lab is taking 24-bit Doppler readings [8 byte doubles, 16 byte long doubles] on simultaneous channels at staggering sampling frequencies, and we can generate a 4GB file in the blink of an eye.

Not to mention MPEGs of e.g. The Ten Commandments or Gone with the Wind.

Re:64-bit Array Indices?

[Golthar]

How is this interesting?
If you run into this limit you obviously need to subdevide the workload.

Make multiple array's to fill the data with.

Re:64-bit Array Indices?

[bellings]

What combination of compiler, language, platform, and operating system currently allow 64-bit array indices for long doubles?

I mean, when you're doing your computations in Fortran 90, or FORTRAN 77, or C (or whatever), what platforms are you currently using to compile and run your programs?

Re:64-bit Array Indices?

[mosel-saar-ruwer]

I mean, when you're doing your computations in Fortran 90, or FORTRAN 77, or C (or whatever), what platforms are you currently using to compile and run your programs?

That's the whole point. It's simply INSANE that we have all this modern hardware but no modern languages to access it.

Writing hacks like doubly indexed 32-bit arrays to access memory above 4GB is just EMBARRASSING in this day and age.

And people have the gall to chide Intel & IBM about A20 address barriers...

Re:64-bit Array Indices?

[Mr. Piddle]

The Java VM can run in full 64-bit mode on SPARCv9 CPUs. Perhaps this would useful to you? Granted, the array index is a limitation for one-dimensional arrays, but perhaps this is an area that could use byte-addressed data streams in a creative way? How about a native call to C-language routines? How about implementing an array-like data structure that overcomes the int limitation? You sound as if you are condemning a good platform for one percieved flaw. Try that strategy on a spouse or significant other and see how far it gets you.

Re:64-bit Array Indices?

[ajagci]

I haven't tried "long doubles" specifically, but GNU C/C++ on Opteron and Linux supports full 64 bit addressing for all other datatypes. If the "long doubles" type is implemented (and I don't see why not), it should work with that, too.

Is there some specific reason why you think there might be a problem?

Re:64-bit Array Indices?

[Anonymous Coward]

Is there some specific reason why you think there might be a problem?

Actually, I've been meaning to submit this as an Ask Slashdot: What is the state of the art of 64-bit programming languages, i.e. which languages/compilers/virtual machines support the following [where "long" is a 64-bit int]?

long [] theArray = new Array[MONSTER_BIG_64-BIT_INT];

for(long i = 0; i < MONSTER_BIG_64-BIT_INT; i++)
{
theArray[i] = foo.bar(i);
}

Inquiring minds wanna know.

PS: As for long doubles, they also offer something of a problem; Intel and Motorola have tended to max out at 12-byte [96-bit] doubles, whereas Sun/Sparc has a 16-byte [128-bit] double.

Again, though, a "write-once, run-any-damned-where-you-please" language/virtual machine is SUPPOSED to handle all of that crap in the background.

If the programmer has to worry about that level of detail, he might as well be writing in assembly language [which seems to be what all good, or at least worthwhile, programming ultimately degenerates into].

Re:64-bit Array Indices?

[Anonymous Coward]

What is the state of the art of 64-bit programming languages, i.e. which languages/compilers/virtual machines support the following [where "long" is a 64-bit int]?

Works fine with C and C++ on Opteron. Probably works fine with Perl and Python as well when they are 64 bit compiled.

Again, though, a "write-once, run-any-damned-where-you-please" language/virtual machine is SUPPOSED to handle all of that crap in the background.

No, it isn't and it never was. You can't haphazardly replace 32 bit types with 64 bit types and expect code to behave the same way. You can't have automatic extension of datatypes depending on the platform and full cross-platform compatibility--it's logically impossible.

Again, though, a "write-once, run-any-damned-where-you-please" language/virtual machine is SUPPOSED to handle all of that crap in the background.

It does. Your numerical code will run out of array indexes on the 64 bit platform in the same way it does on the 32 bit platform. However, it may be able to allocate more than 4G worth of data, which is not quite as good but still nothing to sneeze at.

Re:64-bit Array Indices?

[LionMage]

No idea if Sun has changed the language specification to allow 64-bit array indices, but they seem to have added 64-bit optimizations to some versions of the Java VM, according to these release notes for JDK 1.5. I will refrain from assuming that 64-bit array indices are allowed if you turn 64-bitness on. I noticed, though, that there seems to be a compile time switch that needs to be used for this, not just a run-time switch. This implies byte-code optimizations, at the very least.

Go rent a DVD from Blockbuster.

[mosel-saar-ruwer]

If you run into this limit you obviously need to subdevide the workload. Make multiple array's to fill the data with.

Go grab a random DVD from Blockbuster. There's a real good chance you'll immediately encounter files of size greater than 4GB.

As I mentioned above, in the medical imaging field, we generate files greater than 4GB in the blink of an eye.

In this day and age, there is simply no excuse whatsoever for any aspect of an "Enterprise" system to lack true 64-bit support. Yes, 32-bit support is nice for backwards compatibility [thank you, AMD], but it's just insane that we don't have a plethora of 64-bit programming languages.

It's like we're stuck in the dark ages, circa 1994.

Re:Go rent a DVD from Blockbuster.

[bellings]

Address limits depend on the Operating System, the Hardware, and the Compiler all working together. I'm suprised you can't find some combination of the three that does what you need.

But, Medical Imaging and DVD playback are NOT enterprise applications.

Re:Go rent a DVD from Blockbuster.

[Anonymous Coward]

Address limits depend on the Operating System, the Hardware, and the Compiler all working together. I'm suprised you can't find some combination of the three that does what you need.

As I mention below, the whole idea of "Java" is supposed to be write-once, run-any-damned-where-you-please.

But, Medical Imaging and DVD playback are NOT enterprise applications.

Uhh, and what about those Enterprise file servers, database servers, and authentication systems that have to store all those terabytes and petabytes of medical images?

PS: On the subject of "Enterprise" systems, you do realize that 2^32 ~ 4 Billion, whereas there are approximately 6 Billion instantiations of Homo Sapiens wandering the planet [don't you]?

condemning a good platform for one percieved flaw

[Anonymous Coward]

You sound as if you are condemning a good platform for one percieved flaw. Try that strategy on a spouse or significant other and see how far it gets you.

Except that it's a monster, jaw-dropping, astounding flaw.

The Java VM can run in full 64-bit mode on SPARCv9 CPUs.

I thought "Java" was supposed to be write-once, run any-damned-where you please.

And you guys think Redmond's marketing department lies out their [collective] ass...

Re:condemning a good platform for one percieved fl

[Mr. Piddle]

Except that it's a monster, jaw-dropping, astounding flaw.

The Java Language Specification explicity states that using longs as an array index is an error. There might be a good reason for this, as there were 64-bit CPUs around when Java was being developed.

Also, when an array gets to a point of literally being 4,000,000,000 elements long, perhaps the application really could use some re-work. What applications need such large one-dimensional structures, anyway? Now that I think about it, it would be pretty easy to create larger arrays, anyway, in the way UNIX inodes allow indirection to access terabytes of data. The performance penalty of the indirection isn't huge.

I thought "Java" was supposed to be write-once, run any-damned-where you please.

It is. However, when you move forward to 64-bit address spaces--and use them--it makes going back to 32-bit a little difficult.

And you guys think Redmond's marketing department lies out their [collective] ass...


Actually, Sun is pretty straight-forward about Java. Usually, the lies come out of the mouths of the people who want to believe them.

J2SE1.5 is now in Beta

[MythMoth]

Subject says it all really, but 1.5 contains some substantial additions to the language, some much needed library additions, and in general is full of goodies:

http://java.sun.com/j2se/1.5.0/index.jsp

Font?

[cerberusss]

Offtopic, sorry, but: what is the font that's used in the bigger, red text "When you look at computer systems, they have this sort of yin-and-yang aspect."??

I can't find it in the stylesheets of the page.