A lot of effort goes into large-scale software development. I believe too many non-software people try to solve the challenge on the wrong level.
It is not a problem of better managing work items at all.
Instead, it is a question of tight coupling vs loose coupling and how systems are being set up and operated in a large enterprise.
Ecosystems are complex systems where all entities within do stuff that influences or directly affects the stuff the other entities do or can do. Ecosystems are not ordered systems. There is nothing that controls behavior or direction of development. Everything that goes on is a reaction to what others do or have evolved into. Constant change is paramount. It is also normal that entities disappear (die).
Enterprises have been traditionally set up like a machine with a few controls to operate it. When it was about a manufacturing shop making bicycles that worked well. A large enterprise offering multiple products and services and employing several thousand people is a hugely complex system and cannot be controlled in the same way as the small shop.
Unfortunately the mindset of the small bicycle manufacturer has created bigger and bigger machines and within those enterprises software systems are created that mirror the machine like organizational structures.
The following are a few ideas.
Every enterprise I have coached at so far is always trying to create long-term project plans based on hope. It is the hope that group X will deliver functionality Y by a certain date that creates a lot of fear and thus drives people more and more into control behavior in a desparate attempt to avoid uncertainty. At some point it becomes a vicious cycle.
The simplest way to avoid uncertainty is to use only what is already in production. It is available, likely runs quite stable and there is infrastructure and operations people to keep it running.
Loose coupling is well-known within the developer community. When any system exposes an API, the services it provides can easily be used in a way not envisioned by the original developers at the time.
A lot of web applications are now being created as a client application running in the browser that talks to an API exposed by the server portion of the whole system. That makes it incredibly easy to reuse functionality in a new context.
The better the technical skills of the developers in an enterprise are, the easier it is for them to make any changes to any piece of software. It’s all about being a true software craftsman.
Good code can be written blazingly fast, if you let experienced people do it. And furthermore, writing the code itself is not the problem at all. The challenge is to design something that can be modified easily. For that you need really good developers.
If you let talented people pursue personal mastery in software development, you will have a lot of very good developers. Just get out of their way and invest into creating an environment that is suitable.
Anyone is allowed to make changes and improvements to any system in the whole enterprise.
Changes need to be given back within a reasonable timeframe to the group owning a system. That is similar to open source development.
There is one particular version control system (git) that makes collective code ownership and giving back changes very convenient. Git has a concept and mechanism called pull request.
On the platform GitHub these are frequently used to accept contributions from developers beyond the core team (see details).
If the code for all systems is owned collectively by the developer community in an enterprise and people are allowed to make modifications to reach their goals, it is required that the variants can go into production.
However, that does not mean that people should be encouraged to operate an infinite number of variants for all eternity. Instead the changes should be pulled in by the owners of a system in a timely fashion and only as an exception should the modifying group be allowed to operate their variant.
To play this game right the concept of real options is helpful.
The smaller a system is - that refers to the size of the code base for it - the easier it is for developers to understand it and make modifications to it.
Previous | 30 Nov 2015 | Next |
This article has been posted to social media sites. There might be comments. Just follow the links:
About me
Hello! My name is Stephan Schwab.
As International Software Development Coach and Consultant I help CEOs and Department Leaders to improve value creation and cohesion within their organization. The outcome will be higher quality, customer delight and more revenue.
Learn about my professional experience since 1986.
Professional Services
I'm fluent in these human languages:
Scrum Pair-Coaching to develop technical competence:
Resources for new clients:
Search
Special Content
Highlights of the Year
Living on planet Earth
Open Source Projects
Stay in touch
My Books
Everything
See a listing of all posts on this site.