Re: advantages or disadvantages of ADA over pascal or modula

["John R. Strohm" <strohm@airmail.net>]

"Alfred Hilscher" <Alfred.Hilscher@t-online.de> wrote in message
news:3E296128.4183B70A@t-online.de...
>
>
> karl bowl schrieb:
> > Assume you have to design an embedded sytem, for instance a
> > movementcontroller for a vehicle. In this way you have to decide for a
> > programming language. Would you prefer Modula2 or Ada95 or anything
else?
> >
> > Nic
>
>
> I have a question to you, too.
>
> Assume you have to program a brake control for car, would you do it in
> Ada? Then I would like to know which tools you use. The systems I know
> are based on microcontrollers HC11/HC12 or 8051. _I_ don't know any
> Ada-compiler for these chips. While there are a lot of C develop
> environments, there are only very few for Modula, and none for Ada.
> Please correct me if I'm wrong.
>
> The next point is: C systems are cheap and even free available.  Modula
> systems are expensive. But - Ada prices are far beyond from good and
> bad. Have you ever heard the prices from Green Hills for example? A
> small company with five to ten people can not buy it. And what I've
> heard about ACT-prices, they seem not be suitable, too. Embedded systems
> in non military environment have not some k-dollars available for
> sw-tools.

Have you ever been exposed to basic engineering economics?

In the United States, with all overhead considered, a good engineer costs
his employer on the rough order of $250,000 per year.  I would be very
surprised if the price was that much different in Germany.  When you are
talking about spending that kind of money on software engineers, trying to
pinch pennies on their tools is idiocy.  What you want is tools that will
improve their productivity, that will let them get more work done per year.

Recall that Pratt & Whitney documented 2x productivity improvement using
Ada.  2x productivity improvement, when an engineer costs $250,000/year,
means that an Ada toolset that costs $250,000 per engineer PAYS FOR ITSELF
IN ONE YEAR.  If two engineers use that toolset, and they both get 2x
productivity improvement, the toolset PAYS FOR ITSELF IN SIX MONTHS.

With ten engineers using the toolset, and only a 10% improvement in
productivity, that $250,000 toolset STILL pays for itself in a year.

This, incidentally, is a significant part of the reason why Symbolics LISP
machines, single-user workstations with high five-digit pricetags, sold
faster than beer at Wurstfest parties.  They paid for themselves in months,
at a time when fully-burdened engineers cost about $100,000 per year.

AND FURTHERMORE.

When you are talking specifically about brake controls for cars, you have to
recognize a few things.  First, you are talking about a safety-critical
system.  Anything less than the absolute best NO-FAIL quality is
unacceptable, because people will get killed AND THEIR SURVIVING RELATIVES
WILL SUE YOUR COMPANY OUT OF EXISTENCE.  Worse, if they can find the
registered professional engineer who signed off on the brake control
software, HE WILL GO TO PRISON for involuntary manslaughter.  The accepted
standard in the United States, last I heard, was around $70,000 per injury
and $300,000 per death.  Second, you are going to amortize your development
costs over hundreds of thousands of cars, so a toolset that costs $100,000
winds up costing PENNIES per car.  Third, when those cars start rolling off
the assembly line, if your brake controls are not ready, it will cost the
car manufacturer a small fortune EVERY DAY your controls are late.

In other words, trying to save a few pennies on software development
toolsets is sheer idiocy.