The Return of Workflows (April)

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.

The Common Workflow Language https://common-workflow-language.github.io/ is an open source effort to create a common way of defining portable, reproducible computational workflows to address the needs of the science community, but could have broad application to business task automation as well.