Microsoft Windows is still the most popular desktop operating system in the world, but open source alternatives (mainly Linux and BSD systems) are gaining a wider user base and the Apple Mac, with it’s undeniable cool factor, is making a decent comeback. A lot of people have a lot of different opinions about this, but for a software developer this means one thing: there are now three main platforms, each with a reasonably large number of users that you might want to devlopment software for. The question is which one do you choose?
The most obvious choice would be Windows because that way your software gets out to more people (especially important if you’re actually selling your software). However that obvious advantage is offset by the fact that there is already a large amount of Windows software and your application might just get lost in the crowd unless it’s very exceptional (or you’re a ruthless marketer). The problem with Mac is that not only is its user-base smaller, and Mac users are generally quite discerning users. Linux and BSDs (which are binary compatible) have a slew of their own problems like different GUI toolkits, different software packaging systems and a general aversion to proprietary software. The upshot is that there are plenty of early adopters and a good application can gather momentum quickly.
You could do some research, weigh the pros and cons and then take the gamble, or you could side step the issue altogether. How? By developing cross-platform. You write your program so that it will run on all of the above without the need for drastic changes. Now the problem with this idea is that all of the platforms are different from each other. A program written for Windows will not run under Linux or Mac OS, unless you use some sort of emulator or virtualization. And most users will not go to so much trouble just to run your program. However, developing cross platform can be done and without too much of a fuss. There are a number of ways to do this, but I’m going to be focussing on two ways that are most efficient for software developers.
You write your program in a compiled language like C or C++ and then just used a compiler specific to the platform you’re interested in to compile your program to a binary for that platform. This way you only have to write your source code once and only have to maintain a single source tree. This also gives you access to platforms that are more esoteric than the three common ones mentioned. The GNU compiler can compile to almost 50 different processor architectures (assuming of course that there is an operating system to actually run your program on). The downside is that you have to avoid using libraries and system calls specific to a platform. This can be something of a platform, especially regarding GUIs.
Here you don’t worry about compiling for different platforms. Rather you write your program in an interpreted language which will run without any change for all platforms for which an interpreter exists. You can use a full-fledged virtual machine like the Java Virtual Machine or you can use a more lightweight interpreter solution like Python or Perl. The obvious restriction with this method is that an interpeter must exist for your target platform (and your users must have the interpreter installed). Howver this shouldn’t be a problem if you stick to the three listed above.
Tomorrow I’ll take a look at ways to implement each of the above methods, explore the potential difficulties of each and take a look at solutions and workarounds.