Swing Dance

According to a post on Hans Muller’s blog, a recent survey reports that “Java Swing is the dominant GUI Toolkit for Northern American developers“. I don’t have anything against Swing but I wonder whether these survey results are particularly notable.

First, the lion’s share of new applications are web-based. Swing has little impact there. Sure, you can run a Swing-based Java applet in a browser window but, sorry, that ship has sailed. Web apps using DHTML, AJAX, Flash, etc. own that world now.

Second, where are all of the Swing-based applications? A few hundred Swing apps are listed on Swing Sightings. But that’s a tiny fraction of the thousands of GUI apps running on Windows, Macintosh, Linux, etc.

Third, it’s unlikely that we’ll see a wholesale rewrite of existing apps into Java/Swing. Even in the open source world, the lion’s share of desktop applications are written in C/C++. For example, despite the fact that more than 80% of OpenOffice.org developers are Sun employees, the code is over 10 million lines of C++ not Java/Swing.

Finally, there’s the whole Swing vs. SWT mess. Swing and SWT are fundamentally different approaches to cross-platform development. Swing emulates platform look and feel while SWT uses native code to leverage platform widgets. Sun backs Swing as a standard part of Java while IBM (and the Eclipse foundation) back SWT.

How did this split happen? Early releases of Swing were, to quote an unnamed source, “Big, buggy and looked like crap”. Swing performance and visuals improved considerably in JDK 1.3 but by that point, SWT was off and running as a fundamental part of Eclipse. And the number of vendors supporting the Eclipse platform has grown substantially since then. Certainly SWT has its detractors but they’re unlikely to convince IBM or the other Eclipse foundation members to switch to Swing. For the foreseeable future, Java GUI development will be split between the two camps.