Re: employment with ada

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

"DPH" <rally2xs@compuserve.com> wrote in message
news:4mo7bvc2n70k6eikm3muu2965nbo3m77ov@4ax.com...
> I've just returned from the Software Technology Conference, a large
> conference and trade show for DoD types, held annually in Salt Lake
> City.  What I saw there leads me to say this:
>
> While Ada is truely a superior language, I believe you should rethink
> getting into it on anything above a hobby basis.
>
> At the conference, one of the 40 minute talks was given by Lockheed
> Martin on the Fate of Ada in the Joint Strike Fighter project.
>
> Starting out by saying that they are all personally Ada zealots, and
> strongly believe the langauge to be superior to anything else around,

I would be very interested in knowing the names of the people involved here.
If they were from Lockheed-Martin Fort Worth, I probably know (or knew) some
of them, and it might answer some questions.

> the company was forced, by business realities, to do their safety
> critical software in the Joint Strike Fighter in a safety critical
> subset of C.
>
> The safety critical subset of C is C with 172 restrictions, augmented
> by a source code analyzer to look out for problems.
>
> Why give up on Ada?  They actually did a study - this isn't just
> someone's personal preference or prejudice.  They found:

I'm writing this in two passes.  I would have enjoyed being there, to ask
some pertinent questions.

> 1) No college in this country is teaching Ada.  There may be some
> qualifiers on that that I don't remember, such as "as a major portion
> of their program", or something like that, but in short there isn't a
> source of new Ada programmers, nor is there likely to be.

Oh, really?  In about 1980, General Dynamics/Fort Worth Division cranked up
the F-16 Multinational Staged Improvement Program, which, among other
things, featured new Operational Flight Programs (OFPs) written in JOVIAL
J73, for MIL-STD-1750A and Zilog Z8002.  At that time, to my recollection,
there was one existing JOVIAL J73 compiler, targeting the 1750A, and it was
nowhere near production quality, and there were NO compilers available
targeting Z8000.  Also, there were no JOVIAL J73 programmers available.  Nor
were there any universities teaching JOVIAL J73.  By your reasoning, General
Dynamics should not have been able to build F-16C/D at all, yet clearly they
did: the airplane has been flying for almost twenty years, and is projected
to remain in service for another twenty.  How do you reconcile your results
with their experience?

> 2) If they hire someone and train them in Ada, and designate them to
> program in Ada, all too often that person thinks to himself, "I'm
> learning a dead langauge, with nowhere to go if this project fails or
> completes" and the next thing you know, that person is in an exit
> interview, looking for a job that will provide "marketable skills."

How often did that reason show up in exit interviews for F-16A/B firecontrol
computer OFP programmers, who were using JOVIAL J3B?  How often did it show
up for F-16C/D programmers, who were using JOVIAL J73?  What happened to all
those JOVIAL J73 programmers when F-16C/D switched to Ada in the mid-1990s?

> 3) The people fleeing Ada are right - there were, at last survey 2
> years ago, 5% Ada jobs.  An informal survey of the latest job market
> puts it at around 1%.

There appears to be an interesting assumption in there.  The assumption
appears to be that Lockheed-Martin does not believe in retaining people who
are proven performers, retooling and retraining them as necessary.  In the
1960s and early 1970s, General Dynamics (and many other companies) did a
massive internal retraining program to convert engineers in other
disciplines into embedded software engineers, because they had a critical
shortage of software engineers when software suddenly became important.

So why, if LM were in fact intent on retaining people, would those people
choose to flee?  Possibly, they perceive that LM is not their employer of
choice?  Possibly, there are human relations problems?

> 4) They projected that they would have to go thru several code
> overhauls to change compilers as Ada compiler providers either went
> out of business, or dropped Ada compilers from their product line.

When General Dynamics bought the JOVIAL J73 compilers for F-16C/D, they
faced this very same set of problems, only worse.  THERE WERE NO
PRODUCTION-QUALITY JOVIAL J73 COMPILERS AVAILABLE.  The ONLY bids to develop
compilers came from small business.  See Judy Edwards and Barry Mowday's
paper on how to buy a compiler from a small business.  As a result, GD
bought the compilers INCLUDING ALL SOURCE CODE AND BUILD MATERIALS, so that
they would continue to be able to build it, on whatever platform.

How does the fact that you are using C change this?  Vendors still go out of
business, they still discontinue products.

> They emphasized, over and over, that they are personally Ada zealots,
> but from a business perspective, Ada for much of the JSF code would be
> a boneheaded business decision.  4% of the operational flight program
> will be in Ada, the remainder in that subset of C.  Program-wide,
> including the support software such as trainers, Ada will acount for
> approx 1%.

Having been at GD/FW during F-16C/D, I am peripherally aware of the risks
involved in airplane software development.  Part of your analysis had to be
a risk assessment, of the relative risk of a Class A mishap caused by a
software defect.  It is well-known from the industry that Ada vs. C by
itself has significant impact on software defect density (Consult Pratt &
Whitney for their experience: you buy engines from them).  What did your
assessment of this factor show, given that the cost of a Class A mishap is
many millions of dollars for an airplane, and about the same for the pilot?

--John