Pulp

I believe it's very likely that a single user call will generate multiple tasks. Think about work that is highly parallelized, the tasking system could make short work of that. Something like an applicability calculation that has many subproblems that can be done in parallel.

So if ^ is the need, how can we facilitate that? What is a good user experience in that area?

A bad user experience is if a user call returns a list of tasks that is really long. Monitoring would be almost as difficult as it could possibly be. You would literally have to query each task status or get them via a list, but in either approach you are checking like a lot of things.

A good user experience would be a way to have all of these tasks thought of as a group of some sort. We had "group tasks" in Pulp2. We could introduce a feature later that would address the motivating need in the first paragraph.

So all ^ is my reasoning to say that I think it would be good if Pulp only ever returned exactly 1 resource w/ the 202 response at GA launch and throughout the Pulp3 line.

+1 to return a single task. For situations with multiple tasks, a parent/group task makes sense.

Thanks for the responses. In terms of the MVP, I am thinking since we only have single tasks so I think we should just make this change:

https://github.com/daviddavis/pulp/commit/aeb30622ae94738f06463f857fb331c4b11d1436

After the MVP, we can design a TaskGroup/ParentTask/etc which extends the Task class and could handle multiple tasks while still conforming to our current task response:

    {
        "_href": "http://localhost:8000/api/v3/tasks/88e44b2e-dc36-406f-a852-9ec4db47d58f/",
        "task_id": "88e44b2e-dc36-406f-a852-9ec4db47d58f"
    }

Or we could add new fields too if needed.

+1 to your commit @daviddavis. I agree w/ that resolution.