Re: Ada 9X objectives

[ryer@inmet.inmet.com]

RE:
/* Written  9:59 pm  Sep 28, 1989 by wtwolfe@hubcap.UUCP in inmet:comp.lang.ada */
/* ---------- "Ada 9X objectives" ---------- */
>  ... (notably multiple inheritance).  
>  ... Unfortunately,
>  compiler maturity is just now reaching levels which make the language
>  attractive as a production tool, and in the meantime the continuing
>  advancement in the theory and practice of programming language design
>  have made the delay very costly to Ada's prospects for widespread
>  acceptance.  
>  ... They cite the recently expressed view that inheritance
>  would result in too great a setback for Ada compiler maturity as
>  an example, saying that inheritance has already been proven to be
>  efficiently implementable and that this is merely a smoke screen
>  for resistance to new concepts.  
>  Bill Wolfe, wtwolfe@hubcap.clemson.edu
/* End of text from inmet:comp.lang.ada */

If it costs $500k each to add a high quality multiple-inheritance type
system to Ada compilers, and there are 15 separate baselines (FE's),
then the cost of the feature (at high quality) is $7.5M.  Will enough
C++ and Eiffel users go out and buy one to cover the investment?  Should
the DOD:

   A.  Require a full class/inheritance/etc system in Ada 9X AND prohibit
       the use of Eiffel, C++, Simula, etc. in RESEARCH as well as
       in major production programs in order to force an adequate market to 
       be created?

or

   B.  Put in Ada 9X only what the current market is large enough to
       pay for?

or
  
   C.  Put in multiple inheritance and let the quality and availability
       of compilers settle wherever it may?

or

   D.  <fill in the blank>


Please don't take this note as a flame:  I see the Ada 9X contents issue
as being primarily one of economics, not cultural inertia.  I would be
quite interested in comments about how the "won't-use-it-unless-its-
good-and-it-won't-get-good-unless-you-use-it" stalemate can be broken.

I'm hoping the 9X Requirements Team will ask for "enough building blocks
so you can do object oriented programming reasonably well without
revising the entire type system".  

Mike Ryer
(working, but not speaking, for Intermetrics (an Ada, C, and other
language compiler vendor, applications developer, and contract
research company, which believes in using the best available software
engineering technology as much as anyone, but is also a for-profit business.))