From Internet Week > Java > Java On The Desktop: An Idea Whose Time Has Come ... And Gone (Column By David Strom) > September 19, 2003:

But, silly rabbit, Java is for servers, not for desktops. J2EE is a fine application development platform, and that is where Sun rightly should be placing its dollars, coders and talents. Unseating .Net makes a lot more sense and has a chance of success.

Okay, in some senses he's right, but ultimately he's missing the big picture (or his picture is implicitly smaller and he hasn't quite admitted it).

What's not important for Mad Hatter, Sun's desktop OS, is to knock off Windows for the personal computer user. Sun is not making an OS to bundle with what people are picking up at Best Buy -- at least not yet. What Sun absolutely does need to do is have a business workstation client OS that'll readily interface with their server-side, J2EE apps, whether through port 80 the normal way (with Mozilla, included in Mad Hatter) or through Java Web Start, Sun's preferred method of distributing true client applications nowadays.*

The article's author even realizes this without realizing it, shown when he says:

That isn't to say the Mac is perfect, either. Many times, my Web-based apps just don't run on my Mac browsers, and typically it is because someone has decided to pick the Windows-only version of Java or Active X that precludes any Mac users.

What he misses, as any regular reader of freakinname knows, is that .NET programmers often assume IE, never testing on Mozilla, much less IE on Mac, Safari, or some other Mac-specific browser (see the above link why Mozilla is different).

The point being that any Web Forms app or even Java applets and applications ("Write once, test everywhere") isn't going to work just like you'd expect crossplatform. .NET can assume good, compatible clients without much of a loss -- Windows and IE are everywhere. Home users, business users, everywhere. Sun has no such server/client combo in the business workplace, until now.

By putting Mad Hatter together and standardizing the client's interpretation of the Java app (with Web Start already installed & configured on the OS) or JSP application (standardizing on Mozilla instead of IE), Sun finally has a cradle-to-grave approach to delivering server-side, n-tier, business applications. That's why Mad Hatter is so important.

And as a final point, this is just bad thinking (from the article again):

I became completely turned off to Java on the desktop when I saw Scott talk several years ago about how he runs his own Java desktop at work on his own thin-client desktop.

He talks about how the desktop took 20 minutes to start up, etc. Sun's shifted focus now. It's not a Java OS, it's a standard OS with Java preinstalled (futhermore, see what they've done with OEM vendors and JVMs). Java needs to be parasitic. Now they've finally created a host OS that won't take them to court and do its best to scratch them off its back.

* That said, Web Start isn't the answer for truely crossplatform Java applications. And do note that what J2EE needs to fight .NET really isn't truly crossplatform client applications. The problem is that Sun sells Web Start as if it were the way to deliver Java desktop apps xplat to Joe Q. User.

Check out how I deliver The Digest Handler -- you need apps that look native, not downloads and installation for downloads and installations, when you're pitching to the typical PC user (who sometimes puts CDs in the drink holder the wrong way). For Mac OS X, there's a dmg file and one file you copy to your Applications folder. Simple. For Windows, it's a simple exe that the user can put anywhere on the system, and that checks for a valid JVM before running, telling the user where they can download a more recently released one if theirs isn't up to snuff or a JVM doesn't exist on their box at all! Not a perfect solution on Windows, but much less esoteric than Web Start.

Think the extra installation isn't a big deal? You must be a power user. It is a big deal, and this is the sort of thing Joel's talking about in his article Strategy Letter III: Let Me Go Back!. Web Start throws up about three serious hurdles to installation, losing you about 75% of your possible user base! Take a look at how Excel became king over at Joel's site (link above)