jcwinterton (05/10/82)
Re: Dijkstra's Flames.
WRT Don Chan's article of above title, it is neces-
sary to say that it is true that programming could be more for-
mal. However, I do not subscribe to the idea that programming
is at all mathematical except that some programs do a consider-
able amount of arithmetic. Programming is more analytic logic
than anything else. Also, unless a program specification ex-
ists ists in an inviolable condition, you can never be sure
of correctness. In real world programming, the specification
of a program is often very fluid as the needs of the user
change, and since the user is usually the guy with the purse
strings, you do what he wants. Careless, non-modular program-
ming can make this into a real mare's nest, and this has fi-
nally been recognized by the academic community. Programming as
a science will never be as rigorous as one of the physical
sciences simply because the objects are generally fluid. More
work on formal methods of approach is definitely needed,
but we must all be careful not to go chasing off into the wild
blue yonder trying to find the programming language to end all
programming languages that can produce verified programs which
run reliably as soon as they will pass to the code genera-
tion phase of the compiler. Many programs never produce
"correct" results. Often, it is sufficient to produce acceptable
results.
People who denigrate COBOL, PL/1 and FORTRAN are denying
the inevitable. These languages are here to stay, and nothing
will replace them in the foreseeable future. The reason is very
practical, and it is investment. Millions of dollars have
been spent writing programs in these languages and, due to gen-
eral conservatism, will remain in these languages. At one
time, I worked for a company that capitlaized their programs as
assets and depreciated them as if they were equipment or
furniture. It took a long time to convince them that this was
poor practice, as the life-expectancy of any given version of
a program appears to be somewhere between three and nine months.
Another point about the "commercial" language detrac-
tors is that very few of them have ever written a serious pro-
gram in any of the three except, perhaps, FORTRAN. It is
only natural that anyone who has had exposure to ALGOL based
languages would learn to hate FORTRAN with a passion. Lumping
COBOL into the same class is, however, a gross error. COBOL
was designed to complement FORTRAN (used by engineers and
scientists) for use by the business world for accounting pur-
poses. It has served this purpose well and faithfully, turning
most major commercial computers into printing presses. COBOL
does, however, do something that many of the "nice" languages do
not do, and that is define I/O very carefully and with a real ef-
fort at being useful.
John Winterton.