TaskDetailsStore about exorcist

[dwhswenson]

Planning to switch from the load_task model to the run_task model. Difference in usage:

# load_task usage
task = task_store.load_task(taskid)
result = task()

# run_task usage
task_details = task_store.load_task_details(taskid)
result = task_store.run_task(task_details)

The concern is that the environment where the task runs may be changing over time (e.g., files needed to define the task may change). The load_task model risks that everything is baked into the task when it is loaded, and the task could theoretically hang around a bit before being used, meaning that the world can change around it.

The task_details are always the same, regardless of serialization state. So using run_task only interfaces with the environment at the moment the task is run.

(Granted, good programming practices will avoid problems in either model, but it seems like a potential footgun to me.)

It is also probably more natural to normal people to write a function that returns a result, rather than a function that returns a function that returns a result. (I'm clearly not normal.)