donald (05/11/82)
I must say the opinions of D.J. Molny and John Winterton are appalling!
It reveals the pervasive attitude that mathematics means evaluating
integrals and inverting matrices. This is by no means what is meant by
"mathematics" in the context of programming. If it makes you feel bet-
ter, substitute "logic" for "mathematics". I think Dijkstra just wants
to put programming on a more formal, logical footing, and if this isn't
"mathematics", I don't know what is.
John makes much of the "fluid" specifications of programs, but I think he
is confusing bad development techniques with prototyping.
In a poorly developed system the specifications change frequently (if the
specifications were ever precise in the first place!), a result of the
user not knowing exactly what he wants. In any context other than software
development this would be intolerable! Try building a skyscraper in which
every other day the required dimensions change.
The proper thing to do is to perform prototyping and experimentation so
that the user finds out what he does or doesn't want in his system, and
once that is properly defined, to go ahead and implement it for real.
*AND* the process of implementation should have a solid formal basis so
that we do it right! That's where the "mathematical" approach comes in.
I might mention that I vehemently detract from the "commercial" languages
(FORTRAN, COBOL, PL/I), yet I have written many so-called "serious" (i.e.
real, commercial, for $$$$) programs in all three. The point is that all
three (with the possible exception of PL/I) are clumsy ("Neanderthal" is
the term used by Andy Tannenbaum) tools which hinder the natural
expression of an algorithm. These languages fight the attempts of the
programmer to properly abstract data, modularize programs, and
in general produce good programs.
I might also mention that COBOL I/O is grossly system-dependent and not
as nice as John suggests.
FORTRAN and COBOL are no more inevitable than the horse and buggy.
After all, at the turn of the century there was a considerable
investment in the latter.
Consider the development of OS/360 as a programming project and the
construction of the World Trade Center as an engineering project. What
would we have said if the World Trade Center fell down several times a
week? Yet people fatalistically accepted the analogous behaviour of OS/360.
That is why engineering is a science and programming is a black art.
Until we can construct programs with the same degree of reliabilty as
buildings, it will remain a black art.
Don Chan, utcsrgv!donaldhwtroup (05/13/82)
Myself, I always thought logic was a brnch of philosophy! ...!decvax!utzoo!utcsrgv!hwtroup
drd (05/14/82)
"engineering is a science" !?