The fourth edition of the Domain-Driven Design Europe 2019 conference was held early this year from Jan 31-Feb 1 at Amsterdam. Eric Evans, who is known for his book Domain-Driven Design: Tackling Complexity in Software kick-started the conference with a great talk titled “Language in Context”. In his keynote, Evans explained some key Domain-driven design concepts including subdomains, context maps, and bounded context. He introduced some new concepts as well including bubble context, quaint context, patch on patch context, and more. He further talked about the relationship between the bounded context and microservices.
Want to learn domain-driven design concepts in a practical way? Check out our book, Hands-On Domain-Driven Design with .NET Core by Alexey Zimarev.
This book will guide you in involving business stakeholders when choosing the software you are planning to build for them. By figuring out the temporal nature of behavior-driven domain models, you will be able to build leaner, more agile, and modular systems.
What is a bounded context?
Domain-driven design is a software development approach that focuses on the business domain or the subject area. To solve problems related to that domain, we create domain models which are abstractions describing selected aspects of a domain. The terminology and concepts related to these models only make sense within a context. In domain-driven design, this is called bounded context.
Bounded context is one of the most important concepts in domain-driven design. Evans explained that bounded context is basically a boundary where we eliminate any kind of ambiguity. It is a part of the software where particular terms, definitions, and rules apply in a consistent way. Another important property of the bounded context is that a developer and other people in the team should be able to easily see that “boundary.” They should know whether they are inside or outside of the boundary.
Within this bounded context, we have a canonical context in which we explore different domain models, refine our language and develop ubiquitous language, and try to focus on the core domain. Evans says that though this is a very “tidy” way of creating software, this is not what we see in reality. “Nothing is that tidy! Certainly, none of the large software systems that I have ever been involved with,” he says.
He further added that though the concept of bounded context has grabbed the interest of many within the community, it is often “misinterpreted.” Evans has noticed that teams often confuse between bounded context and subdomain. The reason behind this confusion is that in an “ideal” scenario they should coincide. Also, large corporations are known for reorganizations leading to changes in processes and responsibilities. This could result in two teams having to work in the same bounded contexts with an increased risk of ending up with a “big ball of mud.”
The different ways of describing bounded contexts
In their paper, Big Ball of Mud, Brian Foote and Joseph Yoder describe the big ball of mud as “a haphazardly structured, sprawling, sloppy, duct-tape and baling wire, spaghetti code jungle.” Some of the properties that Evans uses to describe it are incomprehensible interdependencies, inconsistent definitions, incomplete coverage, and risky to change. Needless to say, you would want to avoid the big ball of mud by all accounts. However, if you find yourself in such a situation, Evans says that building the system from the ground up is not an ideal solution. Instead, he suggests going for something called bubble context in which you create a new model that works well next to the already existing models. While the business is run by the big ball of mud, you can do an elegant design within that bubble.
Another context that Evans explained was the mature productive context. It is the part of the software that is producing value but probably is built on concepts in the core domain that are outdated. He explained this particular context with an example of a garden. A “tidy young garden” that has been recently planted looks great, but you do not get much value from it. It is only a few months later when the plants start fruition and you get the harvest. Along similar lines, developers should plant seeds with the goal of creating order, but also embrace the chaotic abundance that comes with a mature system.
Evans coined another term quaint context for a context that one would consider “legacy”. He describes it as an old context that still does useful work but is implemented using old fashioned technology or is not aligned with the current domain vision. Another name he suggests is patch on patch context that also does something useful as it is, but its numerous interdependency “makes change risky and expensive.”
Apart from these, there are many other types of context that we do not explicitly label. When you are establishing a boundary, it is good practice to analyze different subdomains and check the ones that are generic and ones that are specific to the business. Here he introduced the generic subdomain context. “Generic here means something that everybody does or a great range of businesses and so forth do. There’s nothing special about our business and we want to approach this is a conventional way. And to do that the best way I believe is to have a context, a boundary in which we address that problem,” he explains. Another generic context Evans mentioned was generic off the shelf (OTS), which can make setting the boundary easier as you are getting something off the shelf.
Bounded context types in the microservice architecture
Evans sees microservices as the biggest opportunity and risks the software engineering community has had in a long time. Looking at the hype around microservices it is tempting to jump on the bandwagon, but Evans suggests that it is important to see the capabilities microservices provide us to meet the needs of the business.
A common misconception people have is that microservices are bounded context, which Evans calls oversimplification. He further shared four kinds of context that involve microservices:
The first one is service internal that describes how a service actually works. Evans believes that this is the type of context that people think of when they say microservice is a bounded context. In this context, a service is isolated from other services and handled by an autonomous team. Though this definitely fits the definition of a bounded context, it is not the only aspect of microservices, Evans notes. If we only use this type, we would end up with a bunch of services that don’t know how to interact with each other.
API of Service
The API of service context describes how a service talks to other services. In this context as well, an API is built by an autonomous team and anyone consuming their API is required to conform to them. This implies that all the development decisions are pretty much dictated by the data flow direction, however, Evans think there are other alternatives. Highly influential groups may create an API that other teams must conform to irrespective of the direction data is flowing.
Cluster of codesigned services
The cluster of codesigned services context refers to the cluster of services designed in close collaboration. Here, the bounded context consists of a cluster of services designed to work with each other to accomplish some tasks. Evans remarks that the internals of the individual services could be very different from the models used in the API.
The final type is interchange context. According to Evans, the interaction between services must also be modeled. The model will describe messages and definitions to use when services interact with other services. He further notes that there are no services in this context as it is all about messages, schemas, and protocols.
How legacy systems can participate in microservices architecture
Coming back to legacy systems and how they can participate in a microservices environment, Evans introduced a new concept called Exposed Legacy Asset. He suggests creating an interface that looks like a microservice and interacts with other microservices, but internally interacts with a legacy system. This will help us avoid corrupting the new microservices built and also keeps us from having to change the legacy system.
In the end, looking back at 15 years of his book, Domain-Driven Design, he said that we now may need a new definition of domain-driven design. A challenge that he sees is how tight this definition should be. He believes that a definition should share a common vision and language, but also be flexible enough to encourage innovation and improvement. He doesn’t want the domain-driven design to become a club of happy members. He instead hopes for an intellectually honest community of practitioners who are “open to the possibility of being wrong about things.” If you tried to take the domain-driven design route and you failed at some point, it is important to question and reexamine.
Finally, he summarized by defining domain-driven design as a set of guiding principles and heuristics. The key principles are focussing on the core domain, exploring models in a creative collaboration of domain experts and software experts, and speaking a ubiquitous language within a bounded context.
If you want to put these and other domain-driven design principles into practice, grab a copy of our book, Hands-On Domain-Driven Design with .NET Core by Alexey Zimarev.
This book will help you discover and resolve domain complexity together with business stakeholders and avoid common pitfalls when creating the domain model. You will further study the concept of bounded context and aggregate, and much more.