Gary Kildall, PL/I Compiler, versus COBOL
Patrick Logan's questions inspired this comment (slightly expanded since I posted it on Patrick's weblog):
Gary Kildall (wp) wrote one or more PL/I (subset) compilers for micros (8080 and, later, 8086). Later, PL/M was a descendent -- a smaller, simpler language but with somewhat similar syntax -- heavily used at Digital Research on the Operating Systems that were CP/M's descendents. Search "Gary Kildall" "PL/I compiler" will yield pointers to start some research, if desired.
Seems to me that PL/I lost out to COBOL for many of the same reasons that Ada lost out to C++ fifteen to twenty years later. PL/I was seen as this huge, complex beast of a language, with everyone's favorite kitchen sink thrown in, while COBOL was already pervasive, a necessary evil. COBOL's complexities didn't seem to matter as much, because they were already familiar.
PL/I had some funky features that were seen as difficult to implement at the time -- e.g., "on conditions", an early form of exceptions; and arbitrary user-specified data/word-size precision, with a zillion rules about how differently-sized variables would combine.
(Historical notes: Some of my first real programs were written in PL/I, and run on one or the other of the IBM-360/91's at UCLA, around 1968. Also, I worked at Digital Research from 1984 to 1986, the most amusing aspect of which was that they still considered themselves to be serious competition for Microsoft. Finally, I worked at IBM in 2005, and saw some evidence that PL/I was still in use that recently, in some parts of that enormous corporation.)