(This page is now superseded by DMTN-055: Github | LTD)


(from a discussion between Jim Bosch and Gregory Dubois-Felsmann  )

  • Overview
  • Task/Config context and assumptions
  • Functional design / usage pattern
    • Covers the concept that SuperTasks define their own data-grouping rules
    • Specification of inputs and outputs
  • SuperTask class interface
  • Pipeline class interface
  • DataID-mapping model
  • Pre-flight environment (in particular, the design and behavior that's common across all the implementations)
    • “Science DAG” definition
    • using the DataID-mapping tool to implement defineQuanta
    • logic to produce the “Science DAG” from calls to defineQuanta
  • Quantum-execution environment (in particular, the design and behavior that's common across all the implementations)
  • Notes on specific expected implementations of the Pre-flight and Quantum-execution environments
    • CmdLineActivator
    • DRP production
    • SUIT / Firefly / Science Platform Portal Aspect use of SuperTask
    • (open to adding others)
  • Consequential requirements on Butler to support SuperTask (and description of how Butler is expected to be used in the SuperTask framework)
  • Worked examples
    • ISR
    • Coaddition

Things to remember to include somewhere (may get their own section):

  • Points of extensibility vs. things that are constrained by the design
  • What is actually provided in the core SuperTask/Pipeline/Activator* framework, and who is responsible, vs. what is an abstraction/template to be filled in by developers of SuperTasks and platform-specific Activator code


 * "Activator" here is taken to mean all the centrally-provided tooling for building the Pre-flight and Quantum-execution environments