With the advent of mobile computing and (almost) always available network access we should think about the way we are building our applications. We often think in terms of client and server applications. In my opinion we should expand on this and start building “schizophrenic” applications that are clients and servers at the same time even if they are running on the same machine with only one available client at first. Let me elaborate on this:
Exposing an API and thus acting as a “server” is important for many applications to make integration into other software systems possible. It is obvious in the web where you often integrate widgets or services onto your page, for example like-buttons, maps, avatars and so on. It lets others use your software in their programs via scripting and possibly other means. All that broadens the use cases for your application. It becomes more valuable because there are more possibilities for your clients or others to use it. Often your application acts as a client to different services providing value of its own. So it serves two purposes and provides double or even multiplied value!
I would like to make some more specific examples:
- Many vendors of some piece of hardware provide a user program to their hardware as a windows application. There is no (easy) way to remote control the hardware or to use it on different platforms. Sometimes you get a driver library and can build a service around it but it usually takes a lot of work.
- Applications provide only one interface, usually a platform specific GUI and you are stuck with it. Would it not be nice to see specialized views for your mobile device of choice? In the future it is perhaps your brand new smart watch where you would like to see the status.
If your application is build from ground up with other software as a client in mind, you (or others) can add new ways of interacting with your application and the value it provides. This comes with a side-effect that should not be underestimated: Exposing an API helps your design!
You do not only open your application to a plethora of use cases but you will also build a better software. Thinking of the boundaries of the system, designing an API, using it in different scenarios and with different front ends will make you system better structured and much clearer separated into modules. When you create the possibility for different clients of your system you remove most of the danger of mixing UI code with business logic. If your clients have to use a defined API, which is not necessarily the same for all clients, they have to depend on the specified behaviour exposed as a service. It does not matter if the API is Java, CORBA, REST/JSON, SOAP or what not, the pure existence will have you define boundaries to your system. Your application will become part of one or more other systems forcing you to put thought into modularisation and separation at a larger scale than classes or packages. All this will help you with the design and overall your application will handle more use-cases than you might have imagined and will be prepared for changes in computing unforeseen and yet to come.
Adding new frontends or APIs and exchanging different parts of the system become comparatively easy in contrast to many conventionally built monolithic applications.