Irfan Alibay

I'll say the `no-deps` thing is really an initial thought for now, I really need to dig into how big a problem this is. I'm a bit concerned about how...

Looks like just letting pip resolve the missing packages during src build may solve the problem here - at least it doesn't appear to be pulling down MDAnalysis itself.

An initial curveball here - `dask` is a rather heavy dependency, could all this be achieved with `dask` as an optional dependency?

## My 2 cents What I would prefer is to: 1. Avoid class proliferation where possible 2. Avoid the issue where this still occurs because we have various backends available...

> Given that backends are “just” implementation details I really don’t think that user code should need to track what happens in the background. As a user I want to...

> if an algorithm can safely be run in parallel with e.g. `multiprocessing` then I would assume it should also be safe to run it in parallel with dask or...

@orbeckst if I'm understanding this correctly, you're saying that the onus on ensuring that all backends work for all methods / future subclasses, etc.. is squarely on the MDA core...

> Either our ParallelExecutor._compute_with_FRAMEWORK() works for everything or it works for nothing and hence I don't believe that developers should need to care about it and thus ultimately the only...

> We can't guarantee that it will work for code that we didn't write but I think that's the same for every piece of code that we publish. I think...

Thanks for summarising the conversation @marinegor - would organising a call to discuss this all further be wise here? I feel like there's a lot here and it might be...