NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/lang/ada/3760

Internet Message Format | 1992-12-20 | 5.5 KB

Path: sparky!uunet!munnari.oz.au!goanna!ok
From: [e-mail] (Richard A. O'Keefe)
Newsgroups: comp.lang.ada
Subject: Re: C++ vs. Ada -- Is Ada loosing?
Message-ID: <[e-mail]>
Date: 18 Dec 92 04:06:02 GMT
References: <[e-mail]^<[e-mail]> <[e-mail]>
Organization: Comp Sci, RMIT, Melbourne, Australia
Lines: 91

In article <[e-mail]>, [e-mail] (news) writes:
> The more I think about it, Ada (safe, sane....) reminds me of the Volvo.
> You see all those advertisements for Volvos tauting safety features, and
> they never mention the best safety feature, i.e. the fact that they usually
> don't run and you can only hurt yourself so badly sitting under a tree
> in a car.

I have deleted the rest of Holden's flame.  Let me offer a data point.
Just recentrly, I posted an article to comp.lang.c in which I listed a
number of errors in a textbook.  The textbook is about data structures
in C.  The authors understand data structures very well.
    Author X "is chair of the CS dept and Professor of CS&EE at
    <famous university>.  [He] has published 30 proceedings, managed
    research grants totalling over US$3.5 million, ... "
    Author Y "is Professor of CS in <not so famous university but I've
    heard of it even here>.  [He] has published over 140 articles and
    has written several texts.  He is an editor of four professional
    journals and has managed over US$8 million in research grants."
    Author Z "is chair of <a third good university>'s CS dept and a
    veteran of 14 years of teaching."
The authors thank Profesor P, Dr Q, and Messrs R & S for "debugging"
the book.  They also thank Mr T of AT&T Bell Labs (who has written
several books on C himself) for reviewing the manuscript.

Sounds good, no?  The authors also say that the code in the book is not
specially typeset because every example has been tested, and they didn't
want to introduce any errors.

On p6 they write "all programs in this book follow the rules of ANSI C".
The first example is on p7, facing this claim, and it contains an OBVIOUS
violation of the ANSI standard, and a known portability problem (which
X3J11 invented a new macro for the express purpose of fixing), and calls
an operation which may fail without checking whether it succeeded.  The
last example in the book has a type error, and the authors' only comment
is to say that the compiler's error message can be ignored.

In my posting in comp.lang.c I described (with page numbers) several other
mistakes.  My personal favourite is one which appears to reveal a fairly
basic misunderstanding about C scope rules. There are others which should
never have been accepted by any C compiler, although they are legal C++.

Many of the problems are potentially serious.  As it happens, with the
exception of the scope problem, GCC or an ANSI lint _would_ catch all of
them.  But Professors X, Y, and Z "tested" their programs under an awesome
total of two compilers, which evidently did _not_ catch these errors.

I can never quite remember the line in "Life, the Universe, and Everything."
"Don't worry, Arthur Dent ............... be very very frightened."

You see, I had been talking with a publisher's representative, and had
complained about another publisher's recent book which had about one
factual error every two pages.  The publisher's representative said, "well,
OURS are checked.  I'll send you one I think you'll be able to use."
Well, no, I can't.  I think my students are better off being shown how to
write CORRECT C code for a smaller variety of data structures than
INCORRECT C code for an impressive range of data structures.  I also think
that a textbook that TALKS so much about abstract data types should teach
students to DO it, but although it is easy to define a stack type in C
so that a program using it has no clue about the implementation, the code
in this book just isn't like that.  The authors never even mention how to
use header files to get modular code.

Am I worried about C, when experienced people who are managing huge
amounts of money (once upon I time I paid US taxes, some of that was
my money) manifestly don't understand C, even though they imagine they
think they understand it well enough to instruct others?
Worried?  No.

I'm very very frightened.

I believe that Ada is a bigger language than C.  Both languages have "dark
corners".  But there seems to be a tradition of Ada compilers being picky
and C compilers letting it all hang out.  With all due respect to "The
Emperor's New Clothes", which I loved when it came out, I am now _more_
frightened about critical software being written in C or C++ than in Ada.

This is actually an important point.  A programming language is more than
a standard.  There is also a tradition, a way of approaching the language.
(It's something I have observed about Prolog:  if you plug into the
Edinburgh _tradition_, Prolog is a simple, comprehensible, and powerful
tool, but if you just look at a manual, most people never learn how to
use it effectively, and that most especially applies to the authors of
textbooks about it.)  It would be perfectly possible to have an "Ada"
compiler which was as sloppy as a C compiler, and it is possible to have
C systems which are very strict.

My current impression is that Ada textbooks tend to be more accurate in
the claims they make about what is or is not valid Ada, and tend to have
a higher level view of the software process, than C books.  Is this an
illusion, due to my knowing C relatively better than I know Ada?