The Old Beamline GUIs Are Holding Back General Pipelines

Why do most beamlines insist on asking me where I want to save my data? The beamline already expects me, knows what proposal I'm on, and knows where the last thousand datasets went. It's a small complaint, but I think it belies a bigger problem.

Most beamline control GUIs were designed around a person sitting in front of a workstation running a single experiment. At least on non-MX beamlines. Move a motor. Open a shutter. Collect data. Save. Repeat. Repeat again. There is nothing inherently wrong with this, as it worked so well for so long. I just don't think we can bring this into Gen 4 synchrotrons.

The good news is that many new beamlines are no longer running individual experiments; they're running pipelines and series of experiments. Lots of repetition, and the experiment is often the shortest part of the entire workflow.

Yet we still ask users questions the beamline already knows the answer to. Some of these questions made sense twenty years ago. Some even made sense five years ago. Software companies and manufacturers figured out the goal isn't to give users choices. The goal is to make the right thing happen automatically. The best systems are predictable, where self-similar inputs are treated similarly by the pipeline and produce similar outcomes.

Reproducibility is a foundation of science and a relatively new feature at many beamlines. Robots and pipelines are better, storage is cheap(ish), and Covid normalized remote access. We know far more about how users actually interact with facilities than we did a decade ago, and I think that's a big win.

High time we revisit some of the old assumptions and make sure we are putting humans in the correct part of the experiment cycle. Less remembering where files belong, more time designing, validating, improving workflows, and hunting down mistakes before they propagate.

If you could delete one pointless ritual from facility life, what would it be?