Re: From extended Pascals to Ada 95 guide

[Richard Riehle <laoXhai@ix.netcom.com>]

Marco van de Voort wrote:

> >As for submarines, you're right. Since C/C++ is said to replace Ada in US embedded
> >systems.. glups...

Can you document this?  I know of some cases where irresponsible or ignorant systems
engineers
have specified C++ instead of Ada, but there continues to be a lot of Ada in many of
the systems I know about.

Anyone who would select C++ instead of Ada for a safety-critical military weapon
systems either
1) does not understand Ada, 2) does not understand C++, 3) does not understand either
one, or
4) has some agenda that supercedes reliability and technical excellence.   I frequently
encounter
all four situations in my software practice.

There may be some alternative emerging within the C family of languages:  C#.  I
recently attended
a briefing on that language.   It may improve on the reliability issues so rampant in
C++ and rectify
some of the concerns relative to Java.   Of course, it is still a new language offering
and has a
job of persuading real software developers to adopt it.

So far, though, no programming language quite measures up to Ada when reliability and
safety
concerns are the key issues.   I personally like Smalltalk and Eiffel, but would not
choose either
when developing something as important as a military system, a communications
satellite, a
cardiac control unit, flight control systems, or the many other kinds of software that,
if there is
an accident, can result in the death or maiming of unintended humans.

Modula-2 and Modula-3 fall short of the requirement.  C++ falls far short of the
requirement.
Oberon is a step in the right direction.   Eiffel, as a language, has potential for the
future but the
implementations are not at the right level of maturity.   Java is an interesting
offering, but more
of a research topic in many respects.  It is also a good teaching tool. Perhaps it will
improve with time.

When one studies this issue really carefully, with the goal of reliability and safety
in mind, the decision in favor
of Ada naturally falls out as the right one.  It is an engineering decision, not a
programming language decision.
Very few software developers, and even fewer software development managers, are trained
to consider
engineering issues in constructing software.  If one is designing for one of the
popular desktop operating
systems, many of which are inherently defect-infested themselves, there may be minimal
benefit in
selecting a language focused on error-free software.   In that case, C++, Java,
Smalltalk and other popular
languages are fun to use.   I wonder how many C++ programmers would trust their lives
to flying in
an airplane programmed in that language.

Well, this is just Richard preaching to choir again.   Sorry for the rant.   It is the
end of a long day and I just
reviewed this thread, discovering it failed to focus on what is and is not important --
at least to my mind.

Richard Riehle