The Permanent Temporary

There is a catalog of astronomical objects called the New General Catalog. It was compiled in 1888 by a Danish astronomer named J. L. E. Dreyer, consolidating observations from the Herschel family and others into a single reference. It is neither new nor general anymore. It is 138 years old and covers a narrow slice of what we now know exists. Every astronomer alive is aware of this. And yet: NGC 188, a star cluster near the north celestial pole, is still called NGC 188. Not by any quality of the cluster itself—just its position in a list made by one man in the nineteenth century.

You've done this too. You have a folder called "temp" that's been there for two years. A project called "v2" that's now on its eighth iteration. A Slack channel called #launch-blockers that became a general-purpose discussion space sometime in 2023 and never recovered its original meaning. The name no longer describes the thing. Everyone knows it. No one changes it.

I was reading a technical essay about a well-known database system that became famous for its simplicity—it did one thing exceptionally well. Over the years, it accumulated hashes, lists, streams, sorted sets, time-series, JSON support. Each addition partially solved a problem the others couldn't. None of them retired the need for the next one. Now, in 2026, there's a proposal to add yet another data type—an array—because despite having six different structures that approximate arrays, none of them quite get there. Each was a reasonable addition at the time. Together, they've created something the original designers would barely recognize.

What both of these stories have in common isn't sloppiness. It's that naming is infrastructure, and infrastructure has a gravity that accuracy doesn't.

When you name something, you're not just labeling it. You're building a coordinate system around it. Other things start referencing that name—documentation, APIs, mental models, conversations, habits. The NGC numbers appear in thousands of research papers, star charts, and software libraries. Renaming them would mean updating every reference ever made. The database's data types are embedded in client libraries, configuration files, and production systems serving millions of requests per second. You can't just rename "sorted sets" to something more precise without breaking every tutorial, every Stack Overflow answer, every engineer's internal map.

So the inaccurate name persists. Not because anyone defends it as accurate, but because the cost of renaming is real, distributed, and immediate—while the cost of the inaccuracy is vague, cumulative, and borne by newcomers who weren't in the room when the name was chosen.

The people who named the New General Catalog probably didn't expect it to last a century. They were solving a problem they could see: too many disconnected observation lists, no way to know if two astronomers were looking at the same object. "New" and "General" were selling points, not permanent descriptions. The same way "temp" was supposed to last an afternoon, and "v2" was supposed to be the final version.

What interests me isn't that names go stale—it's that the staleness becomes structural. The wrong name doesn't just sit there. It shapes how people think about the thing. If you call a folder "archive," people hesitate to delete anything in it. If you call a channel "urgent," everything posted there feels like it needs attention. The name creates affordances that outlast the intention behind it.

There's a version of this that's optimistic, too. The NGC, for all its anachronism, works. Astronomers can still find things. The coordinate system, though wrong in its self-description, is stable enough to be useful. Maybe there's something to learn from that—not that inaccuracy is fine, but that a shared frame of reference, even a slightly bent one, has value that precision alone can't replicate. A renaming that's perfectly accurate but adopted by no one is worse than an old name everyone already knows how to use.

Still. Every time I open that "temp" folder, I think about NGC 188—a seven-billion-year-old star cluster, named by a number in a "new" catalog from the nineteenth century, sitting near the pole of a sky that doesn't care what we call it. The cluster was there before the name. It'll be there after. The name is just the handle we attached, already rusted, still holding.

Why is it that the systems we design to organize the world always seem to outlive the assumptions they were built on—but never outlive the names?