> OpenAPI schemas replace type-checkers, docker compose replaces service factories, Kubernetes replaces the event loop. Every call across components acrues failure modes, requires a slow march through (de)serialisation libraries, a long trek through the kernel’s scheduler.
This has nothing to do with OOP, and can be made out of structured-programming components, or even purely functional components. In fact, stateless services are a staple of horizontal scaling, and could be a poster child of FP taking over the real world (along with React).
What made OOP problematic was mostly shared and concealed mutable state, and the ill-conceived idea of inheritance. Both of these traits are being actively eschewed in most modern software: mutable state is largely separated into databases, inheritance is often rejected in favor of composition. These are all practical, non-ideological choices, ways to relieve well-known pains. It this regard, OOP is on its way out, even in strongholds where it's ingrained into the very fabric, like JVM languages.
The complaints of the author are mostly about component-based architecture, with limited compile-time trust (know thy vendor) and large run-time distrust (every network request should be seen as a potentially malicious request). This is the price that we pay for having so many ready-made building blocks to choose from, and for the ability to make our services available (or access someone else's services) anywhere on the planet, 24/7.
Highly scalable architectures with their complexity were invented by companies who needed them, like, well, Google or Amazon. There is no good way to serve billions of requests daily. If you serve mere millions, you may not even need all that, and make do with a few beefier boxes. It has nothing to do with OOP, again.
This has nothing to do with OOP, and can be made out of structured-programming components, or even purely functional components. In fact, stateless services are a staple of horizontal scaling, and could be a poster child of FP taking over the real world (along with React).
What made OOP problematic was mostly shared and concealed mutable state, and the ill-conceived idea of inheritance. Both of these traits are being actively eschewed in most modern software: mutable state is largely separated into databases, inheritance is often rejected in favor of composition. These are all practical, non-ideological choices, ways to relieve well-known pains. It this regard, OOP is on its way out, even in strongholds where it's ingrained into the very fabric, like JVM languages.
The complaints of the author are mostly about component-based architecture, with limited compile-time trust (know thy vendor) and large run-time distrust (every network request should be seen as a potentially malicious request). This is the price that we pay for having so many ready-made building blocks to choose from, and for the ability to make our services available (or access someone else's services) anywhere on the planet, 24/7.
Highly scalable architectures with their complexity were invented by companies who needed them, like, well, Google or Amazon. There is no good way to serve billions of requests daily. If you serve mere millions, you may not even need all that, and make do with a few beefier boxes. It has nothing to do with OOP, again.