Re: Is the "Ada mandate" being reconsidered?

[gwinn@res.ray.com (Joe Gwinn)]

In article <nhbui2uhwc.fsf@paralysys>, nasser@apldbio.com (Nasser Abbasi) wrote:
>    Gnu Ada95 is *very* young, both in years and in miles traveled, and I can
>    ^^^^^^^^^^^^^^^^^^^^^^^^^
> Well, Ada95 is older than Java, yet it seems the whole market is
> jumping over itself to use Java. Then how do explain that? I don't see
> any one saying Java is too young lets wait few years untill the bugs
> are out of it befor we use it ? They use it today with bugs and all.  

Ahh, I didn't bring Java up as a vision of maturity.  It's mostly hype
right now, although the claim is that Java (a simplified C++) will push
C++ aside, but I don't know; Java is currently an immature slug.  Nor
would I think of using Java in a military mission-critical system, let
alone to control a weapon.  I accept no responsibility for the Java hype. 
Complain to Sun, et al.


>    Note that the C/C++ world is from ten to one hundred times larger than the
>    Ada world, and had a 10-year head start.  It's not obvious that Ada,
>    however perfect it may be, will ever catch up, because the C/C++
>    "industrial-strength cashflow" is larger than the Ada cashflow by a like
>    ratio.  The rich always get richer.  It's a matter of market size and
>                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>    economics, not technology.
>    ^^^^^^^^^^^^^^^^^^^^^^^^^^
> 
> May be software ENGINEERING is different from other engineering. But in
> other engineering fields one would choose the best technology for the
> job, not the one that has the most market size. 
> This is like a civil engineer choosing plastic to build a bridge instead of
> steel becasue plastic is more popular and has more market share. 

Wrong analogy.  Tell me, what would you say to an engineer who insisted on
designing and fabricating his own pipefittings, rather than going down to
the hardware store and buying what he needed there?  It's not that he
couldn't make better pipefittings than the hardware store can sell him. 
But the hardware-store fittings are made by the tens of millions, work
quite well, have no surprises, and are dirt cheap.  And the cashflow from
such large-scale manufacturing pays for a great deal of research into the
most minute of problems and issues, as major competing pipefitting
manufacturers claw at each other for market share.  No operation not on a
like scale can hope to keep up, never mind overtake.  The fact that the
engineer's fittings are much better is irrelevant.  And, they do make pipe
and pipefittings of plastic.

And, do you disagree with my "from ten to one hundred times larger"
estimate?  This is the key issue, I think.


> Plus, I don't see why C++ has 10 years start over Ada, after all Ada is
> allready an ANSI and ISO standard, while C++ is neither still, offcourse
> this assumes that standards are important thing to look at, May be for
> some people they are not important, but I think they are.

What I said was that "C/C++" had the ten-year headstart, not C++ by
itself.  This is clearly true. The natural upgrade path for C is go to
C++, as one can compile legacy "Classic C" using a C++ compiler.  Many of
the vendors that claim that their product is now implemented in C++ mean
only that their ancient (K&R) C code is now compiling under C++, but that
all the new features are turned off.

I don't spend too much sleep worrying about C++ not being a standard.  It
will be, soon enough, and most of the C/C++ world doesn't care all that
much anyway.  When they complain about the lack of a C++ (or any other)
standard, the issue is either that they didn't get their pet rock into it
just yet, or that portability of source code from one version to another
doesn't work all that well.  True enough.  The primary reason for all this
is that the C/C++ community has not yet decided what to include, and what
to ditch; this decision will be based on field experience, and a lot of
loud arguments, rather than a panel of graybeards.

That said, on (for instance) FAA jobs, where the allowed languages now
range all the way from ANSI C to C++, I have chosen ANSI C, for its
maturity and stability, even when object-oriented designs are required. 
The FAA is *very* conservative.  (Note that X/motif is OO, but not C++.)  

FAA jobs.  After the AAS disaster, the FAA won't hear of Ada.  I know a
number of people of varying levels that worked on AAS, and Ada was far
from the main cause, but it was blackened anyway.  I would guess that all
those loud claims that Ada could solve all software problems boomeranged. 
It shouldn't have been a surprise -- no language, however good, can solve
solve system engineering and design problems.  It's only a language.  An
immature language (or its tool suites) can cause more than its fair share
of trouble, however.  

So, we are back to maturity versus risk versus benefit.  

It always comes down to this, doesn't it?


Joe Gwinn