Changing Paradigms … Pull Principle for the Product?

Oh no, we cannot do that! Why? You can always challenge paradigms. Let us challenge configuration management.

Complexity kills speed. In classical systems we often need to manage variants and versions of components. The combination of both could be called a configuration. 

• Variant means “coexistence of two solutions at the same time”. 
• Version means “temporal succession of implementations”. 

In a cyber-physical system with embedded components, both aspects have to be managed in the classical world. If I want to manage the problem, I need to know all possible implementations beforehand, that is I need to know all configurations of the system and need to validate each of them. At least the ones I want to “allow”. Or limit the number of possible configurations to those that “make sense” (for me) and show the desired functionality and at least validate this subset. I think we can all agree that configuration management is a complex task with a lot of dependencies. 

If managing something is complex and tedious, maybe there is a different solution. (This is a reference to my old atomic physics professor, observing us calculating quantum theory wave functions for an exercise on end-less printer paper. “Are you sure you are on the right path? There must be a simple solution. This is not elegant.” He was right. There was a simple one. I learned later.) Remember: There is always an elegant version of a brute force solution. 

What is the solution when classical management is becoming too slow and complex? Bingo. 

Selfmanagement. Trust the experts/developers/worker a.k.a. value creators. Provide the right objective and guidance, then they are going to figure out the rest. Objective and guidance, a.k.a. a reference frame requires commitment and boldness.

What if we reverse the problem of configuration management and stop managing? Instead of pushing solutions to a cyber-physical system, we let the system PULL the valid configuration? Not used configurations do not need to be validated and tested. The system knows how it is built and can do the job of the “system integrator”. This is simple and fast. It comes at a price tag, but suddenly a new configuration does not add to the bill. Where is break-even?

Not possible and all phantasy? I do not think so. I believe it can be done and I am convinced it is implemented somewhere already. In the automotive world. All you need is an automatic test and release system built into your cyber-physical whatever and a suitable description of the components and their relations. Heard of a digital twin before? I used to think of it as a digital simulation of the system of interest. Possible, but more powerful is “just” a description file, describing your system of interest possibly as a graph.
No guarantee though that is really exists. I am just reading tea leaves here and there. Correlating random information. 😉I would bet on it.

Oh, and one more thing: in mathematics, elegant proofs sometimes are visible when you watch the system from the outside. 

Cheers.