Re: Why don't large companies use Ada?

[beckwb@ois.com (R. William Beckwith)]

Jean D. Ichbiah (ichbiah@jdi.tiac.net) wrote:

: Can you seriously talk of rewriting frameworks of classes as encompassing
: as Borland's OWL, Microsoft Foundation classes, or Zinc? Does not make much
: sense to me and goes against the most important lesson we have learnt in
: the past twenty years:  Reuse, Reuse, Reuse, ...

Reuse is for projects not for products.

Projects typically have low reuse.  COTS software companies have
very high reuse if you count their customers.  If you are writing
software that will be used in only a few projects then reuse is
critical.  However, COTS software gets used in hundreds or thousands
of projects.  Development cost is insignificant to the potential
benefit to the customer base.

Thus, most software companies partially or completely rewrite their
products every three to five years.  Look at the most visible examples:

    * Oracle did it with version 7.
    * Sybase did it with System 10.
    * Rational did it with Apex.
    * Borland did it with version 2.0 of OWL.
    * Microsoft did it with MFC before they even released the first
      version! (see OOPSLA proceedings a couple of years ago for MFC
      design team article)
    * We at OIS are on our second re-write of Screen Machine.

Certain bodies of code are so frequently reused by others that it
is imperative to make fundamental, architectural improvements to
the code to increase flexibility, reliability, and performance.

The products you mention, OWL, MFC, and Zinc, are fairly thin object
oriented covers above a non-object oriented API.  They are much better
interfaces that the underlying API's.  However, they are not even
close to the depth and power of a fully implemented class framework.

The real frameworks of the future don't use the underlying toolkits
(Taligent, OpenStep, Fresco).  They provide a much broader, more
flexible architecture that is implemented in that O-O language.

Again, I must mention the Smalltalk communities `lack of reuse'.
They don't reuse Motif or (most of) MS-Windows.  (I see everything
through GUI glasses ;-)

: Not only the overwhelming difficulty of rewriting the code.  But you will 
: have to test it, certify it, and ... maintain it.  And if you have not been
: discouraged by this mountain of effort, you will have to produce user
: documentation of comparable quality - in paper an on-line - to that 
: existing for these frameworks.

So I guess you believe that David Weller should be writing bindings
to the C++ Booch components instead of re-writing them?  Uggh.
Makes me shudder at the thought.

We have a better language than the developers of the C++ frameworks did.
We will end up with better frameworks by using the languages power, not
by binding to other frameworks.

: See you in a few hundred man-years.

I'd like our customers to think that it takes us `a few hundred man-years'
to write our products.  But in practice, the development, testing, and 
maintenance of bindings closely rivals the re-implementation of a
framework using more native approaches.  And the bindings can't even
compare to the functionality of the pure approach.

: Jean D. Ichbiah

... Bill Beckwith