#Deependee

##Language philosophy

• everything is a dependency
  • fields depend on computations
  • functions depend on implementation
  • dependency graphs (traces) are always visible
• dependencies can be existing or latent
• arrays and objects are streamed by default
• no invalid states
  • no null values
  • no exceptions

##Platform philosophy

• cloud-ready
• code/config based deployment and integration
• no logging
• observable
• versioned and time series
• automated input validation

##Environment philosophy

• data storage is cheap
• networks are fast but
  • need to consider geolocation
  • network distribution needs to be obvious
• machines can make better decisions
  • heuristics, coded best practices, algorithms
  • machine learning
• small compute is cheap and ubiquitous
  • high threading counts
  • lambdas
  • on demand compute
• as a result of these, with time
  • performance increases instead of decreasing
  • TCO decreases instead of increasing

##Testing

• when a constraint is defined before assigning a dependency, the assignment is checked against the constraint
• when a constraint is defined after assigning a dependency, the constraint is checked against the value
• tests can probably be executed both ways, but the standard way is to define constraints after dependencies

##Design notes

• Parser -> automatic instantiation
• Registry -> multiple implementations possible, in particular for connecting to other registries
• Deployer -> target dependent
• Executor -> multiple strategies, each with their own implementation
  • resolves dependencies based on strategies
    • local direct call
    • local thread pipe
    • remote simple call
    • remote call with retry/circuit breaker
    • remote call with throttling
  • communicates with other executors to resolve upstream and downstream dependencies
  • highly multithreaded
  • recalc: https://lord.io/blog/2020/spreadsheets/