Also mentioned in the linked tutorial: Memory on early mini-computers was limited, sometimes only a few thousand characters. Mumps programs were loaded into memory as source code rather than as compiled binary. This was done because it was determined that compiled Mumps programs would be far larger than the corresponding source code versions, especially if the source code employed size reducing abbreviations.

So why didn't they use a bytecode interpreter, or at least a minifier (as is done with JavaScript), so that the programmer could keep the source code readable while still giving the interpreter a compact version?

Would it have been too computationally expensive, or had people just not come up with the idea yet?