One of the realities of application development is that there are a lot of factors that go into its underlying architecture. I came of age during the Epoch of Client-Server Architecture (also known as the 1990s and later the .com era) and we were taught, very firmly, that despite the implication of "client-server" that there were but two, distinct "tiers" comprising an application, there were actually three. The presentation layer, the GUI, resided on the client. All business logic resided on the server, and a third, data access tier completed the trifecta.
This worked well because developers tailored each application to a specific business purpose. Thus, implementation of business logic occurred along with application logic. There was no real reason to separate it out.
As applications grew in complexity and use, SOA came of age. SOA introduced the principles of reuse we still adhere to today, and the idea that business logic should be consistent and shared across all applications that might need it. I know, revolutionary, wasn't it?
Before SOA could complete its goal of world domination, however, the Internet and deployment models changed. Dramatically.
Today, an "application" is a no longer defined necessarily by business function, but by a unique combination of client and business function. It's not just the business logic, it's the delivery and presentation mechanism that make it unique and of course, more challenging for operations. Business logic has moved into a converged business logic-data tier more commonly known as The API. The client is still, of course, responsible for the presentation.
But application logic - the domain of state and access - is in flux for some. As illustrated by the ill-advised decision to place application-logic in the presentation layer, some developers haven't quite adopted the practice of deploying an application (or domain) logic tier to intermediate and maintain consistent application behavior.
But there always exist unique processing that must occur based on context. In some cases, some data might be marked cacheable as a means to achieve better performance for mobile clients when communicating over a mobile network while in others, it might not be because the client is a web-based application running on a PC over a LAN. A native mobile application deals with the state of the user "are they logged in" differently than a web application relying on cookies. The business logic should not be impacted by this. Ultimately the business logic for retrieving order X for customer Y does not inherently change based on client characteristics. It cannot, in fact, be shared - reused - if it contains application specific details regarding the validation of state, unless such an implementation uses a lot of conditional statements (that must be modified every time a new method is introduced, by the way) to determine whether a user is logged in or not.
Thus, we move the application-specific, the domain, logic to its own tier, usually implemented by or on a proxy that intermediates between the API and the client.
Domain Logic Tier
Business Logic Tier
• Maintains an internally consistent model representation on both sides of the app (client and server)
• Is ontological
• Often involves application state and access requirements such as "the user must be logged-in and an admin" or "this object is cacheable"
• Consists of elements and functions that are specific to this application
• Concerned with coordinating valid interactions between presentation and data (client and data)
• Is teleological
• Often involves direct access to data or application of business requirements such as “if order is > 1000 apply discount X”
• Consists of elements that are common to all applications, i.e. does not rely on a given UI or interface
What's most fascinating about this change is that a "proxy" tier traditionally proxies for the server-side application, but in this new model it is proxying for the client.
That's not odd for developers, because if you break down the traditional model the "server" piece of the three-tiered architecture really was just a big application-specific proxy to the data. But it is odd for operations, because the new model takes advantage of a converged application-network proxy that is capable of performing tasks like load balancing and authentication and caching as well as mediating for an API, which may include transformative and translative functions.
IMPACT ON DEVOPS
What this ultimately means for devops is an increasing role in application architecture, from inception to production. It means devops will need to go beyond concerns of web performance or application deployment lifecycle management and into the realm of domain logic implementation and deployment. That may mean a new breed of developer; one that is still focused on development but does so in a primarily operational environment, in the network.
It means enterprise architects will need to extend their view into operations, into the network, and codify for developers the lines of demarcation between domain and business logic. It means a very interesting new application model that basically adopts the premise of application delivery but adds domain logic to its catalog of services.
It means devops is going to get even more interesting as more applications adopt this new, three-tiered architecture.