#eavoices

While Gartner seem to be the primary advocates of software defined everything, (SDx) it’s rather obvious that SDx is primarily focused on service delivery, infrastructure and networks. I give full credit to Jason Bloomberg for exploring software defined development and devops in his recent blog. [1] He merges SDx and Shift Left ideas saying,  “If we combine no-code with DevOps properly, we now have a way of abstractly representing working production software, including its functionality. Not just the limited-scope apps that some no-code platforms are best known for, but full-blown, enterprise-class applications – created from nothing but their abstract representations with the push of a button.”

Heady stuff! In fact Jason  dismisses any skepticism advocating high risk strategies as necessary for digital transformation, “Given the exponential pace of technology innovation all around us, the greater risk is that we miss the full significance of Software-Defined Everything and its impact on the digital enterprise – until it’s too late.”

Wise heads may be shaking at this point; many will remember early attempts at diagrams to code, computer assisted systems engineering, model based development etc. All of which work in a limited manner, but never achieved the level of capability that is necessary to support enterprise class environments. However it has been clear for some time that this is a realistic goal. The basic concepts were articulated by Jack Greenfield and Keith Short in 2004 while they were at Microsoft, described as the Software Factory – “the configuration of extensible tools, processes and content using a software factory template based on a software factory schema to automate the development and maintenance of variants of an archetypal product by adapting and configuring framework based components.” [2]

Where I part company with Jason is in his promotion of “no-code” as the primary ambition in conjunction with integrated DevOps. The software factory is a “low” code environment, unlike no-code environments under full control of the developer community, who codify common patterns to allow standardization and assembly of common code to perform repeatable tasks; establishing an exceptionally high quality and productivity environment customized for the unique needs of a specific enterprise. But the really big value adds are that the factory governs reference architecture compliance and facilitates evolutionary development, enabling continuous change in initial development and the operational state, because all artifacts are under management. That’s not the full story; there are some key principles that make this practical:
1. A formal reference architecture is required to allow full life cycle meta data management and standardization of delivered artifacts. Note here I mean reference as meta-objects. A key aspect of the reference architecture is that everything is delivered as a service according to defined behavioral and dependency policies that enforce separation of concerns at all levels and reduce dependencies.
2. Rigorous specification of business services and rules independent of implementation. (the abstract representations referred to above) that support test driven development and provide high visibility of intra-project dependencies.
3. Abstracted specification interface removes requirement for (UML or similar) modelling skills. Allows business or business proxies to be responsible for the specification task.
4. Design by contract establishes strong boundaries, visible dependencies and separation of concerns.
5. Automation of infrastructure and codification of exemplar patterns in common code using software factory concepts.
6. Late (generation time) bound technology, allowing open technology architecture (platforms and frameworks) and inheritance of architectural changes with minimal impact on functional code.
7. Design by exception – only new patterns require design. Regular patterns are already codified in the factory.
8. Seamless linkage with Continuous Integration, Test Automation and DevOps.
9. Scalable Agile delivery process uses managed artifacts to provide high visibility of dependencies.
10. Continuous modernization process enables evolutionary capability development and change.

Reading this list, you might be tempted to say, this is just model driven development, or limited code environment, or SOA code generation etc. But the sum of the parts is greater than the whole – it’s a new model which is inherently business driven, with clear separation of business specification and implementations and technology. We call it the Agile Service Factory. [3] Today it is being used to deliver very large scale, enterprise class modernization projects [4] with exceptional productivity and quality, and the outcomes are inherently agile business services and solutions.

References:
1.  Building the Software-Defined Digital Enterprise (Part 2), Jason Bloomberg
2.   Software Factories, Jack Greenfield and Keith Short, Wiley, 2004
3. The Agile Service Factory
4. White Paper: Modernization with Service Architecture & Engineering and the Agile Service Factory