What is it and Why....

Geode in its current form is the ultimate Monolith. Littered with God objects and tight coupling. This makes maintaining and extending the current Geode system really hard, as changing functionality in ObjectA would most inevitably cause a change in ObjectB, which then cause ObjectC having to change, etc...

The current tight coupling of code makes it harder to reason over the expected behavior of the system, as code is layered on top of code, each layer potentially influencing behavior that is unpredictable.

In order to start unwinding this ball of yarn we need refer back to some Good Design Strategies:

1. Separation of Concerns

2. Modularity

3. Abstraction

4. Anticipation of Change

5. Incremental Development

These strategies are best described in the following links:

• Seven Good Design Strategies Part1
• Seven Good Design Strategies Part2
• Seven Good Design Strategies Part3

Proposal Goals

The proposal specifically targets the Separation of Concerns, Modularity and Abstraction strategies. The proposal will try and address the following goals:

• Classloader Isolation
• Dependency management
• Dependency injection
• Loose Coupling of Modules with high cohesion
• Extensibility and Plug-ability
• Lifecycle management 
  • Bootstrapping
  • Extendible process flows
• Configuration
• Simplified Testing of modules
• Support for Interceptors:
  • Operation Traceability (like Zipkin)
  • Logging
  • Security
  • Transactions
• Event / Message driven
• Clearly defined SPI / API / Annotations
• Open/Close Principle 

In addition to these goals, the solution needs to support:

• Reactive Streams
• Reactive Manifesto
• Back pressure enabled operational flows
• Improved metrics collection

Additional "nice-to-have" features would be:

• Lock-free framework for operations that require a critical section
• Async messaging / Actor models