Re: Legislative Mandate for Ada; mindless translations

Re: Legislative Mandate for Ada; mindless translations

[Charles E Eaker]

In article <2467@sparko.gwu.edu> mfeldman@seas.gwu.edu () writes:
>Is there any consensus on this out there (after all, I'm stuck in the
>Ivory Tower :-))?

Mindless translations are the expected response to mindless mandates,
and mindless mandates appear to be on the rise, especially in
government contracts. So, if code that does the job exists in some
other language, a mindless translator can give you the same algorithms
working in whatever language is mandated for the current project and
save a lot of money. There are enormous incentives to do just that, and
there is no reason to believe that the output of such a translator is
any more or less maintainable than the input.

Actually, some believe that translators can be produced which will
generate code which satisfies coding standards and conventions which
experience has shown to be more readily understood, modifiable, etc.

--
Chuck Eaker / P.O. Box 8, K-1 3C12 / Schenectady, NY 12301 USA
eaker@crd.ge.com        eaker@crdgw1.UUCP       (518) 387-5964

Legislative Mandate for Ada

[Michael Feldman]

I recently received a copy of the section of the Defense Appropriation
Conference Report regarding Ada, and thought you might be interested in
reading what Congress has to say. For you outside-the-Beltway folks,
a conference report is the congressional document that reconciles any
differences between House-passed and Senate-passed bills. Both houses vote 
on the conference report, and basically that's how the law is passed.
In this case, congress passed this DoD appropriation bill at the end
of October, and Bush signed it. Here is the relevant paragraph:

"Sec. 8092. Notwithstanding any other provision of law, after June 1, 1991,
 where cost-effective, all Department of Defense software shall be written
 in the programming language Ada, in the absence of a special exemption
 by an official designated by the Secretary of Defense."

In plain English: no gobbledegook about "embedded systems" or "mission-
critical systems." The criterion is cost-effectiveness. Might be fun to
chat on the net about how big a loophole "cost-effectiveness" is, or
how it might be determined.

As background, here is a lengthy paragraph from the explanatory language
that came along with the conference report.

"Ada Programming Language - The Department of Defense developed Ada to
reduce the cost of development and support of software systems written in
the hundreds of languages used by the DoD through the early 1980's.
Beside the training economies of scale arising from a common language,
Ada enables software cost reduction in several other ways: (1) its
constructs have been chosen to be building blocks for disciplined
software engineering; (2) its internal checking inhibits errors in
large systems lying beyond the feasibility of manual checking; and
(3) its separation of software module interfaces from their
implementations facilitates and encourages reuse of already-built
and tested program parts. While each of these advantages is important,
Ada's encouragement of software engineering is fundamental. Software
practitioners increasingly believe the application of engineering
disciplines is the only currently-feasible avenue toward controlling
unbridled software cost escalation in ever-larger and more complex systems.
In march, 1987, the Deputy Secretary of Defense mandated use of Ada in
DoD weapons systems and strongly recommended it for other DoD
applications. This mandate has stimulated the development of commercially-
available Ada compilers and support tools that are fully responsive to 
almost all DoD requirements. However, there are still too many other
languages being used in the DoD, and thus the cost benefits of Ada are
being substantially delayed. Therefore, the Committee [congressional
conference committee - MBF] has included a new general provision,
Section 8084 [changed later to 8092 - MBF] that enforces the DoD
policy to make Ada mandatory. It will remove any doubt of full DoD
transition to Ada, particularly in other than weapons systems
applications. It will stimulate DoD to move forward quickly with 
Ada-based software engineering education and cataloguing/reuse systems.
In addition, U.S. [government] and commercial users have already
expanded tremendously the use of Ada and Ada-related technology.
The DoD, by extending its Ada mandate, can leverage off these commercial
advantages. Navy Ada is considered to be the same as Ada for the purposes
of this legislation [HUH? What's Navy Ada? Anyone know?], and the term
Ada is otherwise defined by ANSI/MIL-STD-1815. The Committee envisions
that the Office of the Secretary of Defense will administer the general
provision in a manner that prevents disruption to weapon systems that
are well into development. The Committee directs that applications
using or currently planning to use the Enhanced Modular Signal Processor
(EMSP) be exempted from mandatory use of Ada as a matter of policy."

This is what is known as "legislative history." It is not formally
part of the law but gives insight into the mindset of the lawmakers
(or their staff people, really). Have fun with it.

Mike Feldman

Re: Legislative Mandate for Ada

[Bruce Benson]

In article <2449@sparko.gwu.edu> mfeldman@seas.gwu.edu (Michael Feldman) writes

>"Sec. 8092. Notwithstanding any other provision of law, after June 1, 1991,
> where cost-effective, all Department of Defense software shall be written
> in the programming language Ada, in the absence of a special exemption
> by an official designated by the Secretary of Defense."

"shall be written" - does this mean new software or does it mean we have
to convert the hundreds of millions of lines of Cobol to Ada?  We may be
able to do it blindly with a Cobol to Ada translationr and it would
probably be fairly cheap (as things go) to do so.  I can see the metric
the bean counters are going to use to check compliance:

                          Total KLOC - Ada KLOC
                          ---------------------
                              Total KLOC

If the percentage doesn't approach 100 fast enough then they will mandate
mindless translations.

>"Ada Programming Language - The Department of Defense developed Ada to
>reduce the cost of development and support of software systems written in
>the hundreds of languages used by the DoD through the early 1980's.
>Beside the training economies of scale arising from a common language,
>Ada enables software cost reduction in several other ways: (1) its
>constructs have been chosen to be building blocks for disciplined
>software engineering; (2) its internal checking inhibits errors in
>large systems lying beyond the feasibility of manual checking; and
>(3) its separation of software module interfaces from their
>implementations facilitates and encourages reuse of already-built
>and tested program parts. While each of these advantages is important,
>Ada's encouragement of software engineering is fundamental. Software
<on and on on the benefits of Ada>

If I've learned nothing else while working at the SEI, it's that most
software engineering claims are purely back-of-the-envelope
no-connection-to-reality sheer speculation, or in other words:
never been validated on the scale being discussed.  If the government
would simply recognize that their programs are just national experiments, 
and conducted them as such, then we could gain some benefits out of all
the mandated "good ideas" by using government as one big test bed.  This way 
we could justify the high cost of government by reminding everyone that
inefficiency and failure are valid and acceptable results when testing an
hypothesis.


* Bruce Benson                   + Internet  - bwb@sei.cmu.edu +       +
* Software Engineering Institute + Compuserv - 76226,3407      +    >--|>
* Carnegie Mellon University     + Voice     - 412 268 8469    +       +
* Pittsburgh PA 15213-3890       +                             +  US Air Force

Re: Legislative Mandate for Ada

[Michael Feldman]

In article <9700@as0c.sei.cmu.edu> bwb@sei.cmu.edu (Bruce Benson) writes:
>
>"shall be written" - does this mean new software or does it mean we have
>to convert the hundreds of millions of lines of Cobol to Ada?  We may be
Well, I've been reading Ada stuff for about 10 years, and never saw even
a hint that old systems were to be converted to Ada just for the hell of it.
>
>If the percentage doesn't approach 100 fast enough then they will mandate
>mindless translations.
Oh, Lord - I hope not! Seriously, I don't have a lot of inside information,
but I really don't think this is what it's about. Do you have any serious
info suggesting this mindlessness? (I realize that we're all cynics in this
business, but fact and cynicism shouldn't be confused. Got facts?)
>
Mike Feldman

Re: Legislative Mandate for Ada

[Charles H. Sampson]

In article <2455@sparko.gwu.edu> mfeldman@seas.gwu.edu () writes:
>In article <9700@as0c.sei.cmu.edu> bwb@sei.cmu.edu (Bruce Benson) writes:
>>
>>If the percentage doesn't approach 100 fast enough then they will mandate
>>mindless translations.
>Oh, Lord - I hope not! Seriously, I don't have a lot of inside information,
>but I really don't think this is what it's about. Do you have any serious
>info suggesting this mindlessness? (I realize that we're all cynics in this
>business, but fact and cynicism shouldn't be confused. Got facts?)

     I haven't seen any mandate yet, at any level, but expressions of inter-
est in mechanical (read mindless) translation from various languages to Ada
keep cropping up in the part of the Navy that I work with.  Usually some
software house is trying to sell the Navy on the benefits of transitioning
to Ada and the wonderful tool they have to aid that transition.  I've had
to critique a few of these efforts.  The translated Ada was appalling, even
when the proponents were claiming that the program should be maintained (for
decades) in Ada rather than its original, natural, language.  These efforts
always seem to be presented as quck-and-dirty proofs-of-concept, with pro-
mises that the final product will be truly wonderful, as soon as a lot of
our tax money is spent.

                            Charlie

Re: Legislative Mandate for Ada; mindless translations

[Michael Feldman]

Referring to Charlie Sampson's note on translations:

Of course the computing business has seen this translation stuff before.
I hope nobody would seriously consider just mechanically translating
Cobol to Ada for the hell of it. Would you all agree that unless a
system needs _serious_ revision, we shouldn't fix what ain't broke?

Given pragma INTERFACE and some reasonable way to call Ada programs from
another language (I know, it's not easy as things stand now), it seems to
me that even a multi-language system is better than either perpetuating
old languages just for module-to-module compatibility or mindlessly
translating badly-written Cobol into badly-written Ada. Who agrees ?
Is there any consensus on this out there (after all, I'm stuck in the
Ivory Tower :-))?

Mike Feldman

Re: Legislative Mandate for Ada; mindless translations

[Matthias Ulrich Neeracher]

In article <2467@sparko.gwu.edu>, mfeldman@seas.gwu.edu (Michael
Feldman) writes:
> Referring to Charlie Sampson's note on translations:
> 
> Of course the computing business has seen this translation stuff before.
> I hope nobody would seriously consider just mechanically translating
> Cobol to Ada for the hell of it. Would you all agree that unless a
> system needs _serious_ revision, we shouldn't fix what ain't broke?
> 
> Given pragma INTERFACE and some reasonable way to call Ada programs from
> another language (I know, it's not easy as things stand now), it seems to
> me that even a multi-language system is better than either perpetuating
> old languages just for module-to-module compatibility or mindlessly
> translating badly-written Cobol into badly-written Ada. Who agrees ?
> Is there any consensus on this out there (after all, I'm stuck in the
> Ivory Tower :-))?

Amen ! (From another Ivory Tower).

Could it be that sometimes even for new projects, a multi-language system 
is better than a single-language one. When I read about writing expert 
systems in Ada, I have the impression that this is not exactly the kind
of thing Ada is good at. Wouldn't it be better to write the "expert" part
in Prolog ?

> Mike Feldman

Matthias

-----
Matthias Neeracher                                   mneerach@iiic.ethz.ch
   "These days, though, you have to be pretty technical before you can 
    even aspire to crudeness." -- William Gibson, _Johnny Mnemonic_