My first job out of grad school in 1981 was with the University of Wisconsin clinical cancer center. We built a DB to track clinical trial data. I was 1/2 the programming staff and we wrote the code in MUMPS running on a special purpose OS that ran on a Data General minicomputer. The base of our code was a DB-like program written by the VA and distributed all over.
The OS was the truly terrible part. The disk was formatted into 1K blocks and these blocks were the closest things to files that it had. The blocks had names and were organized alphabetically. Programs that didn't fit in 1K had to be broken into multiple blocks, executed in a chain alphabetically.
Mumps had some OK features like good string pattern matching and the array like structures were handy. Each array was the equivalent to a DB table and each row a tuple. No query language though.
To effectively use the ridiculous 1K block structure, all reserved words were 1 character and so where all variable names.
A legal statement was
I I W W
So, as others have said, just writable but un-readable. I gave a talk at a MUMPS conference and was one of very few trained programmers there. Most were medical professionals who did some coding.
Some of the people who had worked on this project before me had gone off to form Epic Systems.
I wouldn't want to code in it again, but it was a good job.