1. Introduction
        describe the purpose of the document
        describe the problem being solved: reconstruction, simulation, trigger
        make it clear that we are not solving the analysis problem

2. Different metrics for optimization
   -- maximize througput
   -- minimize latency

3. Properties of the reco/sim/trigger problem
   These are to distinguish our problem from the analysis problem
   -- use events as a whole (not sure if this is true)
   -- use the object model view of the event data (as opposed to the tabular view common in analysis)
   -- command-line driven (as opposed to interactive)
   -- coordination of large-scale processing (so provenance tracking is valued)
   -- touch full datasets (is this part of large-scale processing?)

4. Requirements
   -- technology constraints (grid, HPC, cloud platforms)
   -- use of computing node resources (memory, coprocessors, storage)
   -- use of networking resources

5. Software engineering issues (desired qualities of our implementation, to help assure we are meeting functional requirements)
    composibility
    configurability
    debugability
    encapsulation-ility
    extensibility
    instrumentability
    lifetime-of-code-ability
    maintainability
    portability
    programmability
    reusability
    scalability
    sustainability
    testability

6. Time periods of interest
   -- HL-LHC era
   -- DUNE startup
   -- Icarus?

7. Granularity of processing
   -- What are the processing elements? Services, tools, algorithms, modules, conditions
   -- What is scheduled?
   -- What is managed?
   -- What are the supported concepts?
   -- Single-pass or multi-pass processing?
   -- What are the customization points of the program? (plug-in models, definitions of data types)

8. Granularity of detector readout data
   -- Is the Event as a universal container old?

9. Configuration
   -- a point of connection with the workflow management system

10. Data context (non-event data, ancillary data)
    -- calibration data
    -- conditions data
    -- geometry
    -- handling anything to do withunderstanding "event" data
    
11. Monitoring program performance

12. Monitoring physics performance
    -- data quality monitoring

13. Modernization
    -- best practices
    -- use of new C++ language features
    -- multi-language programming

14. What high-level primitives ought to be supported?
    -- map-reduce
    -- scatter-gather
    -- stencils

15. Interfaces to external "frameworks" (e.g. visuaization)

16. What other working groups are impacted by our WG?

17. Using pieces of the framework outside of the framework context
    -- address problem of testing individual components