Alexandre E. Kopilovitch wrote:
>
> Roger,
>
>   Probably I'll be unique contributor to this discussion, who agrees that
> it is hard to oppose the arguments presented by you.

I'll agree that it's hard to oppose them, simply because we're talking
about trying to change the preconceptions people hold.  It's easy to
change a person's opinion if he/she realizes that it IS an opinion.

Here, we're trying to change ideas that are just as obviously true from
observation as the fact that the Sun goes around the Earth.


>the gap between strong programmers and poor
> programming workers is much broader and deeper inside Ada community then
> inside C/C++ community. Indeed, that gap in Ada case is easily observable,

I suspect that's the more true version: it's easier to see the
difference between good developers and sloppy hackers when using Ada.

In fact, I've observed a fairly smooth continuum in Ada developers.  I
will grant that the pool is smaller, so it may be harder at a given
instant to find an example of a developer at skill level X.  But I
think the distinction you're seeing is between the incompetent and the
competent, not between the weak and the great.  I see a LOT of competent
Ada developers out there, and a lot of competent software engineers and
developers who could very easily learn Ada.


>   Surely, if you can hire for your projects several really strong programmers
> ...then Project A will have
> all chances to win against Project C for all imaginable criteria. But in
> reality you have a sharp deficit of even moderately-strong Ada programmers
> and at the same time sufficiently good C/C++ programmers are generally
> available.

If you can only get mediocre programmers, go with a mediocre language.

    :-)    <-- smiley added for the humor impaired per ADA

I've worked on a large Ada program (150+ software developers) and
some of them were top-of-the-line.  And then, hoo boy, some of them
were lazy, some were badly trained, and some were just plain stupid.
Between them and me (I hope) were a lot of people in the middle.

One thing I observed is that the better people are wildly more
productive with Ada than with C.  Modularity, typing, generics,
in-language multitasking -- they act as "force multipliers" for the
good people, and if you put them in lead positions they can use these
facets of the language to break the work out for the junior team
members and give THEM the same "force multiplication."

I also observed that mediocre and poor programmers tended to resist
"all that software engineering crapola" and tried to code as if they
were using a fairly old version of FORTRAN.  But if they caved in and
started to really use the Ada language at a design level, they started
getting the same benefits.  And they started liking the language,
because all of a sudden they were getting more productive and were
being viewed as better developers -- not because they were following
some arbitrary code standard, but because they were delivering code
faster and it worked better.


>   As for cost prediction of a project, there are two points to mention:
>
> 1) naturally, there are less chances for big overrun in C/C++ case simply
> because an initial estimate is substantially bigger;

I'd agree that a given overrun is a larger percentage of a smaller
total.  However, overruns tend to scale up with the size of the
project, especially those that hit late in the program -- which are
generally both more common and larger.

This is especially true when the project is back-loaded, since the
code you are throwing away was a bigger part of the original cost.
That is, if you throw out 10% of the code, and coding was 40% of the
project, you have lost 4% of your work.  But if coding was 80% of the
project, you have lost 8% of your work.  And, for the same conceptual
error, you will probably throw out more code -- that's one reason for
a careful, modular design, to "firewall" subsystems so that problems
won't force rework across the entire system.


> 2) Ada projects tend to be optimized. And it is a general law (or at least,
> general experience) that various estimates for optimized things are often
> unstable -- they may jump significantly after small changes.

Here, once again, is the confusion between language and process.

If Project A didn't maintain a management reserve for contingencies,
and an awareness that problems will occur -- not may, but WILL -- and
will require time and manpower to solve, they would blow out their
schedule and budget while using any language.  This is a good lesson
to learn.  Tying it to the use of a specific language is just
scapegoating, and will KEEP you from learning the real lesson.

On the other hand, if Project C wasn't tied down with specific
requirements; if they could deliver something "good enough" whenever
the schedule expired, so long as it resembled the last prototype; then
they could hardly have had any kind of schedule slip or cost overrun.
And, if that's all the customer needs and wants, then that's exactly
what you should do.  There's no point to doing a full set of
architectural drawings for an outhose.

Best,
Sam Mize

--
Samuel Mize -- [log in to unmask] (home email) -- Team Ada
Fight Spam: see http://www.cauce.org/ \\\ Smert Spamonam