Why /usr is called that, and what it was for

1970sEarly Unix disk capacities

The directory called /usr holds programs, libraries and documentation. The name suggests users, and users are exactly what it held.

The machines Unix was written on had disk packs measured in megabytes. The system filled one of them. When it did, the obvious move was to attach a second and put the least essential thing on it, and the least essential thing was everybody's home directory.

1971 onwardsEarly manuals and directory listings

The manuals of the period show it plainly: /usr/ken, /usr/dmr and the rest were people, not packages. A user's files lived under a directory named for them, on the second disk.

laterThe second migration

Home directories moved again, to /home, and /usr kept everything else that had accumulated there in the meantime: binaries, libraries, manuals.

So the modern meaning is the residue of a move. The directory is named for what left it.

1970sWhat the split actually looked like

Mounting a second disk did not mean copying files at random. The system had to boot and repair itself from the first disk alone, so everything needed before the second disk was attached had to stay behind: the shell, the essential tools, the means of mounting.

That division is why the directory structure looks duplicated to this day. Two places for binaries, two for libraries, two for configuration. The duplication is not redundancy; it is the boundary between what the machine needs to stand up and what it needs to be useful.

present dayWhat happened once the reason went

Disks stopped being small, and the boundary stopped meaning anything. Systems now commonly merge the two hierarchies and keep the old paths as links, so that decades of scripts referring to either location keep working.

That is the ordinary end state for the constraints in this part. The reason disappears, the structure remains, and eventually somebody collapses the structure while preserving every name that ever pointed into it. Nothing is removed, because removing a name breaks something nobody can enumerate.

The pattern this is an instance of

A directory named for a purpose, filled with something else, and then explained by an expansion invented afterwards is not unique to this case. The same sequence has produced several of the names in a modern filesystem, and it is why explanations of directory layout so often sound tidier than the history.

The tidiness is the tell. Where a naming scheme is perfectly regular in retrospect and nobody can point to the document that defined it, the regularity was usually imposed later on an arrangement that grew.

present dayWhat to do with a name like this

Nothing, is the honest answer, and that is worth saying plainly because the alternative wastes effort. Renaming would break an unknowable amount of software for an aesthetic gain.

What the history is good for is prediction. Knowing that the boundary between the two hierarchies was about what a machine needed in order to boot tells you which one to look in when something fails early, and that is worth more than the etymology.

present dayWhat the layout is good for now

The one durable use of the old split is diagnostic. When a machine will not finish starting, the question is which hierarchy the missing thing lives in, and the answer tells you whether the fault is in the part that must work unaided or the part that arrives afterwards.

That is a small return on a fifty-year-old accident, and it is larger than the return on knowing the etymology, which is why the entry ends here rather than with the name.

1970sThe same story, one directory along

The directory for temporary files has a comparable history and a different ending. It was a place on the disk that survived reboots by accident rather than by promise, and software came to rely on both properties in different places, which is why its behaviour is specified inconsistently to this day.

The pattern is the same: a location chosen for a practical reason, depended on for properties nobody guaranteed, and then formalised in a standard that had to describe existing behaviour rather than choose it.

1980s onwardsWhat the split cost in practice

Two hierarchies meant every package had to decide which one it belonged to, and the decision was made separately by every distributor for thirty years. The result is the familiar situation where the same program lives in a different place on two machines that are otherwise identical.

Scripts responded by searching a list of locations, which is why the variable holding that list is one of the oldest pieces of configuration still in daily use, and why it is long.

unverified as originThe expansion that came afterwards

The reading of the name as an abbreviation for system resources is widespread and appears in documentation. As an explanation of how the directory came to exist it is a backronym: an expansion invented later to fit a name that already meant something simpler.

It is a reasonable description of what the directory now contains, which is why it spread. It is not where the name came from.

What we cannot verify

The account of the disk filling up is well attested in recollections by people who were there and is consistent with the capacities and the manuals of the period. We have not found a contemporaneous document recording the decision itself, and the specific disk model named in various retellings varies between accounts.