Saturday, July 22, 2006

Can we learn anything new from Swing?

I have done something that all developers should be cautious of; I found one technology I really liked and settled without tracking the competition. The technology I am referring to is SWT! A few years ago I stumbled upon SWT and I haven't looked back since. SWT was faster than Swing, felt better than Swing and of course, my Java apps finally stopped looking like "Java Apps". For years I was happy, I told everyone how much better SWT was than Swing, and I wondered why the whole world hadn't switched over.

Recently I started listening to a few of the Java Posse's Podcasts. The Podcast consists of 4 guys talking about New and Noteworthy things regarding Java. They don't seem to have much experience with Eclipse, although their general Java knowledge seems pretty good. N.B: They are definitely Swing Guys! After listening to their Podcast about Swing Labs I thought I should check it out. It looks like Swing has come a long way in the past 4 years, and the Swing developers credit some of the success to the arrival of SWT. They have claimed that SWT gave them a kick in the rear. They even claim in the SWT vs. Swing battle, Swing has won... I wouldn't go that far!

I don't understand why anyone would want something that almost looks like "Real Widgets!". That is like saying I like my floor because it "Almost looks like Hardwood!" The quote I enjoy most is, "On Windows, Swing looks like windows, and on Linux it looks like Linux..." Looks like Linux!!! What on earth does that mean? Including myself, we have 4 heavy Linux users in our lab, and we all have different Themes, Colours, and Fonts... What does a Linux Button look like?

But when it comes to Emulated (Non-Native) Widgets, Swing may have stumbled upon something. Swing Labs are a group of developers who are writing "Sexy" Swing apps to show the power of Swing. As they develop these applications, new widgets are componentized and they eventually make their way back into the SDK. I was wondering if something like this would be useful for SWT. I am aware of the Nebula project; however, the widgets here still seem to have a business focus. Also, this project is focused on the creation of widgets, not demonstrating their uses through new “toys”. I am primarily talking about applications and widgets for the average home user.

Don't get me wrong, I think native widgets are very important and most applications should be using these. However, for the next Funky I.M. application, Map Viewer, Media Player, Image Browser, Video Game, etc... Fancy looking, Drop Shadowy, Web 2.0ish User Interfaces can have a huge impact. If you don't believe me, just look at the number of skins people have created for WinAMP / XMMS. I personally don't like the look of iTunes, but many people do. It is this type of customization which is very difficult in SWT and it may be something to consider as we move forward.

Does anyone have any thoughts (without getting into the age old SWT vs. Swing battle).

Wednesday, July 05, 2006

96 Votes and Closing!

Finding technology advice using google can sometimes be a real pain. For example, try using google to search for GTK Performance related links (performance tweaks, performance improvements, etc...). Most of the hits are pre 2004, and it is hard to tell if the advice given (or roadmaps drawn out) are valid anymore.

The same thing holds true for Eclipse Bug#37683 (Address platform-specific UI performance problems). Most of the advice here is out-of-date, and many of the problems reported have been fixed. While the SWT team has done a remarkable job over the past few years fixing performance problems, there are still several "Platform Specific Performance Problems" with SWT (Bug #134760, Bug#51693, among others). Steve Northover has proposed closing this bug, and I completely agree with him! This bug report provides very little value to the SWT user and developer communities. It is too general to use for SWT planning and too out dated for the average user. However, the bug does have 1 very valuable contribution, it is one of the most voted for bugs and it indicates that platform specific performance problems are important to the Eclipse community. If you are interested in SWT performance, please voice your opinions about this bug report in bugzilla!

I think comment #211 is an excellent idea. Each release the SWT team could create a bug report that we could use to track performance related issues for that release. Concrete bugs that are currently being investigated could be linked to this report. This would not only give the user community a chance to pitch in on current development efforts, but it would also give us a chance to share up-to-date information regarding performance tips with one anther. For example, I just recently discovered how much Xinerama hurts performance when it is enabled for dual monitor support. Now this may be common knowledge to most people out there, but I bet there is at least 1 other Linux user who did not know this, and it would be great to have an updated performance forum to share this information.

It might even be a good idea to distill the most relevant performance advice and post it to a page on the Eclipse Wiki.