LJ Archive

Don't Code for Linux

Haavard Nord

Issue #106, February 2003

Better to pick the platform after you write the application.

The days of developing applications for a single platform are history. Why? Because every platform offers at least one key benefit that cannot be attained on any other; Windows, Linux/UNIX, Mac OS X, embedded Linux and others each offer unique advantages. But given changing market conditions, it's impossible to predict which platform will give you the competitive edge you need.

Our anwser is: why pick? We feel that developers can and should leverage all the best qualities of each platform by embracing multiplatform development. This is true not only on the desktop, but also on the server, the network, mobile devices and every other tool that connects us. Our increasingly mobile working style demands portable data and portable applications to match today's distributed networks and global organizations.

Organizations that want to compete and survive must recognize a fragmented OS environment as a given, and they must respond by developing applications that run quickly, cleanly and natively on the greatest number of platforms possible. Applications written in this way take advantage of the best features each platform has to offer, without having to be written and rewritten for every instance. This process limits your company and represents a colossal waste of time. Forward-looking companies already recognize that single-platform development is destined to fail, and they have embraced a better way. Here's why we think this report on the death of single-platform development is not exaggerated.

Single-Platform Development Is Expensive

If you wish to develop for more than one platform, thus expanding your target market, your costs rise dramatically. You need a full team to develop for each platform. Perhaps more importantly, you need a full team to maintain and support each platform. This represents a linear increase in cost for each platform—an extremely inefficient way to do business.

Single-Platform Development Closes Doors

Developing applications for one platform increases your risk because you have to choose among markets before their potential is clear. Who's to say you'll be right? Software companies have been made or broken by this choice. In the recent past, people said Windows (with its momentum and market dominance) was the obvious choice—but wait! Linux has proven itself as a serious competitor in the server space and is picking up serious momentum on the desktop and in the embedded space. World-class consumer and enterprise companies are embracing its power, flexibility, security and low cost. So, what used to be an obvious platform decision isn't so obvious anymore. Do you know when (or where) this kind of rapid transformation will happen again? I don't.

Single-Platform Development Does Not Encourage Mobile Deployment

Perhaps most importantly, if you limit development to a single desktop or server platform, you immediately restrict your access to the fastest-growing software market in the world: mobile systems. If, for example, you write an application for Microsoft Windows NT/2000, you automatically eliminate any cost-effective way to run your application on a mobile device, because you have to rewrite the source. Given that it's nearly essential to make applications mobile, developing applications on a single desktop/server platform can be a death sentence for that application even before it's finished.

The software industry has struggled for some time to develop commercially viable strategies for multiplatform development, and its history is littered with companies that have tried to do this, and failed. Why?

One difficulty has been a lack of complete functionality. Many toolkits deliver only subsets of functionality on multiple platforms, not the whole set. Another problem has been reliance on emulation or virtual machines. Both of these impose a significant and usually unacceptable performance penalty, especially for mobile devices that need high performance the most.

It's a well-recognized fact that differences between virtual machines lead to implementation workarounds and tweaks, as well as increased maintenance. This is another expense, and it makes the developers who have to do this work miserable.

Today, though, proven ways exist to write an application once, compile and run it anywhere. Companies who do multiplatform development create an environment in which development innovation will once again be the order of the day—not the exception.

Haavard Nord, cofounder and CEO of Trolltech, started his programming career trying to find acceptable multiplatform toolkits for database development. He now drives Trolltech's efforts in single-source, multiplatform software development. The company's products encourage innovation by letting developers write single-source applications that run natively on Windows, Linux, UNIX, Mac OS X and embedded Linux.

LJ Archive