The Last Known Copy of a Cloud
We speak of data in the cloud as if it is something nebulous and eternal, suspended forever in a digital stratosphere. It is a comforting metaphor, suggesting both ubiquity and permanence. A cloud is always there, seen from anywhere. But a real cloud is a temporary congregation, a fragile consensus of vapor and wind. It forms, it changes, and it dissipates, leaving only a memory of its shape against the sky. This, I’ve come to think, is the more accurate analogy for so much of what we now consider permanent. We are archiving clouds, trying to pin down their ever-shifting edges, knowing the sun will inevitably rise to scatter them.
I encountered this feeling while tracing the lineage of a single, obsolete municipal dataset. It was a simple thing: a live-updating map of public park irrigation schedules, hosted on a city’s "smart infrastructure" portal. For a few years, it was a perfect artifact of its moment—a commitment to transparency and efficiency, built on a specific stack of JavaScript libraries and third-party mapping APIs. Then, a budgetary cycle passed, a vendor contract lapsed, and the portal was modernized. The new site was sleeker, faster, but the map was gone. Not deleted, precisely. Its constituent parts—the data schema, the API endpoints, the visual tiles—were simply allowed to fall out of alignment. The cloud dissipated.
What remains is the last known copy. In this case, it’s a single JSON snapshot captured by a conscientious web archiver, a digital naturalist who saw the clouds gathering and pressed a specimen. Opening the file now is a quiet, technical séance. The data points are all there: coordinates, valve IDs, runtime minutes. But the context that made it a living map—the code that rendered it, the server that authenticated requests, the user manual that explained its color codes—is absent. The data is preserved, yet its function, its *reason for being*, has evaporated. It is less a map and more a fossilized set of instructions for a machine that no longer exists.
This is the quiet tension at the heart of preserving our public, digital present. We are adept at saving the droplets—the final reports, the PDF minutes, the static snapshots. But we are only beginning to learn how to save the weather. The interactivity, the dynamic dependency, the live connection between a citizen and a functioning service—these are the atmospheric conditions of modern public records. To preserve only the output is to save a rainfall total while forgetting the feeling of the storm.
Perhaps the goal is not to believe in the permanent cloud, but to become better, more humble meteorologists of data. To document not just the data points, but the precise pressure systems that gave them meaning: the software versions, the service agreements, the user pathways. To acknowledge that some clouds are meant to be witnessed in their fleeting forms, and that the act of preservation is sometimes less about freezing a thing in place, and more about learning how to describe, in exquisite detail, the shape of something that was always destined to vanish.
Notes & further reading
A few pages I came back to while writing this: