A very good article, touching on many of the problems arising in workflows, also in the scientific field. Interesting with the highlight of differences between

A very good article, touching on many of the problems arising in workflows, also in the scientific field. Interesting with the highlight of differences between "forward" and "backward" defined workflows (let data/tasks flow forward on success vs require some of the most downstream tasks or targets, and let the engine figure out upstream targets), as these have profound implications for how easy some workflow constructs are implemented. E.g. handling processes producing dynamic number of outputs is much easier in the forward defined one.