Re: Customer balks at Ada -- any hope?

[mjsilva@my-deja.com]

In article <8l2pqo$im7$1@nnrp1.deja.com>,
  wv12@my-deja.com wrote:
>
>
> In article <8l01s4$gnr$1@nnrp1.deja.com>,
>   mjsilva@my-deja.com wrote:
> > We're bidding on a custom industrial controller, and I've proposed
to
> > write the firmware in Ada.  The powers-that-be here are satisfied
with
> > that, but the customer is afraid nobody will be around to maintain
it.
> > They're happier with C or C++, alas.  Anybody have any good answers
to
> > their concern?
>
> >
> > I realize that implicit in their position is a belief that Ada
offers
> > no great tangible benefits to the project (even though the machinery
> to
> > be controlled is big, expensive and remotely-located),
>
> C has been known to control big, expensive hardware. One such
> example is the mutinode Deep Blue capable of searching a few million
> nodes per second.

Nowhere did I say otherwise (only a fool would suggest such a thing).
However, I firmly believe, as do most Ada users, that for a given
amount of effort an Ada program will have fewer defects than a C
program.  To argue otherwise goes against all the evidence.  According
to NASA, "The choice of 'C' is to be avoided for our domain of interest
because the language lacks the features that permit robust, reliable
programming."  I have the added datapoint of having determined over
this last year that about 90% of customer-discovered bugs in one of our
current C products would have been caught by Ada before the product got
out the door.

 Is the speed critical in this project? If so, I see
> on reason to avoid Ada that checks every shift, rotate, add, multiply
> in your software.

How much do you actually know about Ada?  A great many checks can be
optimized out, and others can be turned off at the discretion of the
programmer.  I have seen a figure of 7% for "typical" Ada overhead
(that's less than 2 months on the Moore curve).  So as I see it, 7% vs
1/10 the in-field bugs -- it may not be exact, but it sure looks like a
"tangible benefit" to me.
>
> > course strongly disagree with.  As I see it, the arguments are (1)
Ada
> > will offer tangible benefits, both in reliability and in development
> > time, and (2) a decent programmer can pick up similar languages
fairly
> > easily, especially for maintainence.  (Perhaps I should show them
some
> > Ada source...).  Ideas?
> Maybe you could try to sell the safety critical side of Ada. But
> software that does not get tested will crash, kill, dump core, etc...
> (Ariane comes to mind)

Who said anything about not testing software?!  Am I writing in
invisible letters that only you can read?  It's a known metric that for
each additional development phase that a defect persists after its
introduction, the cost of finding and fixing the defect rises by as
much as an order of magnitude.  Bugs caught in testing are much more
expensive to fix than bugs caught at compile time.  I have personally
spent hours or days finding bugs that Ada would have caught in
seconds.  And I *hate* debugging!

BTW, Ariane was not an Ada failure.  It would have come apart exactly
the same if the software had been written in C.
>
> You are not convincing me. Besides, the customer is always right.

I didn't set out to convince you, and you clearly are determined not to
be convinced, so we'll have to call this one a draw.  Why are you so
actively antagonistic?  I have a longtime C background and I happen to
be convinced that Ada is a better solution.  If you have evidence that
this isn't the case then maybe you can try to convince *me*.

Mike


Sent via Deja.com http://www.deja.com/
Before you buy.