A message in computer software refers to a unit of communication that contains information that the source wants to communicate to the outside world, either to specific recipients or as a broadcast to all recipients connected now or in the future. The message itself is nothing, but a container holding information to identify the type of the message and the data associated with it. Messaging is used in a variety of ways. One way is either through the Win16/32 APIs with WM_* messages being sent for any user input or changes occurring in the UI. Another is things affecting the application to XML messages used to integrate systems. It could also be typed messages inside the software, modeled directly as a type. It comes in various forms, but the general purpose is to be able to do it in a decoupled manner that tells other parts that something has happened. The message and its identifier with its payload becomes the contract in which the decoupled systems know about. The two systems would not know about each other.

With your message in place, you want to typically send it. Publish/subscribe, or, as it is known in shorthand "PubSub", is often what you're looking for. The message can be broadcasted and any part of your system can subscribe to the message by type and react to it. This decouples the components in your system by leaving a message for them, and they are unaware that they are direct to each other. This is achieved by having a message box sitting in the middle that all systems know about, which could be a local or global message box, depending on how your model thinks. The message box will then be given message calls, or will activate subscriptions, which are often specific to a message type or identifier.

Over the years, we've had all kinds of message boxes. In Windows, we have something called a message queue, which basically also acts as a message box. In addition, there are things such as service buses that pretty much do the same thing at the core as well. The notion of having something published and something else subscribed is not new.

The message box can be made smarter, which, for instance, could be by persisting all messages going through so that any future subscribers can be told what happened in the past. This is well presented by the following diagram:

A representation of how the subsystems have a direct relationship with a message box, enabling the two systems to be decoupled from each other

There are quite a few paradigms in the art of programming and it all boils down to what is right for you. It's hard to argue what is right or wrong because the success of any paradigm is really hard to measure. Some people like a procedural approach to things where you can read end-to-end how a system is put together, which often leads to a much coupled solution. Solutions are things put together in a sequence and the elements can know about each other. The complete opposite approach would be to completely decouple things and break each problem into its own isolated unit with each unit not knowing about the other. This approach breaks everything down into more manageable units and helps keep complexity down. It really helps in the long term velocity of development and also how you can grow the functionality. In fact, it also helps with taking things out if one discovers one has features that aren't being used. By decoupling software and putting things in isolation and even sprinkle some SOLID on top of this (which is known as a collection of principles; this being the Single Responsibility Principle, Open/Closed Principle, Liskov Substitution Principle, Interface Segregation Principle, and Dependency Inversion Principle).

By applying these practices with decoupling in mind, we can:

The ability to publish messages rather than calling concrete implementations becomes vital. These become the contracts within your system.

This book will constantly remind you of one thing: users are a big part in making this decision. Making your system flexible and more maintainable is of interest to your users. The turnaround time to fix bugs along with delivering new features is very much in their interest. One of the things I see a lot in projects is that we tend to try to define everything way too early and often upfront of development, taking an end-to-end design approach. This often leads to overthinking and often coupling, making it harder to change later on when we know more. By making exactly what is asked for and not trying to be too creative and add things that could be nice to have, and then really thinking of small problems and rather compose it back together, the chance of success is bigger and also easier to maintain and change. Having said this, decoupling is, ironically enough, tightly coupled with the SOLID principles along with other principles to really accomplish this. For instance, take the S in SOLID. This represents the Single Responsibility Principle; it governs that a single unit should not do more than one thing. A unit can go all the way down to a method. Breaking up things into more tangible units, rather than huge unreadable units, makes your code more flexible and more readable.