Re: Ichibah flames, and flames out over, Ada 9X

[agate!usenet.ins.cwru.edu!magnus.acs.ohio-state.edu!cis.ohio-state.edu!ne ws.sei.cmu.edu!ajpo.sei.cmu.edu!wellerd@ucbvax.Berkeley.EDU (David Weller)]

In article <1993Feb24.211028.8076@evb.com> jgg@evb.com (John Goodsen) writes:
>How true.  A technological feasible solution in no way gaurantees
>a success.  There is a concept of "perceived" need versus "actual"
>need.  You don't market to "actual" needs, but to "perceived" needs.
>The perception in the OO community is that the class is the natural
>structuring concept.  The Ada 9X approach to classes, while technically
>feasible is going to have a tough sell into NEW markets, because it
>lacks the perception necessary to be embraced by a mainstream of
>OO developers.  In particular, the concept of a "class" as a basic
>object oriented structuring mechanism.  This same argument also
>applies to the lack of direct support for multiple inheritance in Ada9X.  
>While there have been proposals on how to perform MI using generics
>or method delagation (which are still debatable), the perception 
>of multiple inheritance as a core component of the language is not
>there and will also hinder the marketing of Ada into NEW markets ...
>While those of us who already know and like Ada are excited about
>9X features, these 2 lacking perceptual needs (Class level packaging
>and Multiple Inheritance) will negatively impact the growth potential
>of this powerful language.
>
>Before the flame war on this starts again, execute the following:
>
>
> If Flamer Has_Not_Read(DRAGOON) then
>    Redirect_Flames ( to => "/dev/null" );
> end if;
>
>--
>John Goodsen
>Software Process & Environments
>EVB Software Engineering
>jgg@evb.com

Well, since I have not only read DRAGOON (actually, the _book_ has a
different title :-), but am also working with Colin Atkinson on 
extensions to DRAGOON (and translation thereof into Ada 9X), I feel
qualified to respond to Mr. Goodsen's points.  WARNING: This won't be
a flame -- hope I didn't disappoint anybody. :-)

The 9X community, IMHO, has struggled with this problem, that the
_perception_ that 9X will be a "weaker" language than C++ because
it lacks MI and is not class-based.  Indeed, this was the crux of my
argument against tagged types many months ago (which lead Jean Ichbiah
to chime in with me on this issue, which lead to the tagged vs.
class debate, which lead to "The Vote", which lead to Mr. Ichbiah's
resignation as a DR, which lead to a GCN article on it, which lead to
my reply here, which created this thread.  Ain't it funny how things
come back to haunt you?).  Since that time, I've had several frank
discussions with Tucker, who has done a commendable job of creating
an extension to Ada which supports features that are necessary for
Good Software Engineering(tm).

Ada 9X will indeed have a "tough sell" to the unwashed masses who
believe that MI is a Good Thing for an OO language.  Adding MI
was discussed at great length on here a couple of years ago, culminating
with TUcker's survey of OO developers (trying to find where MI
was actively/genuinely used in _real_ OO projects).  The survey
revealed little, if any use.  And, as I recall, many places indicated
they forbid the use of MI (lattice effects tucked in a class four levels
away are HORRIBLE to fix!).  After careful consideration, the 9X MRT
decided that MI was too difficult to add into Ada SAFELY without adding
excessive overhead.  I'm dredging this all up from memory.  Perhaps
Tuck or Ed Schonberg can throw some more comments in here.  

As for Ada not being class-based -- GREAT!  I'm convinced that, in the
long run (which Ada was designed for, right? :-), tagged types will
prove FAR more versatile.  The problem is SELLING the concept to people
who are so closed-minded as to believe that class-based languages are
the "bee's knees" for OO (then again, one could say us "taggers" are
close-minded, huh?).  Mind you, I was NOT an easy convert on this issue.
I raised my share of stink on the 9X mailgroup (look back in Aug/Sept 92
timeframe).

I'd be interested to hear what your arguments are for making Ada
class-based (meaning, I haven't heard a NEW one for months).  Not
that it matters anymore anyway.

Oh, one final sales pitch for DRAGOON: Seems that the Italian company
that markets DRAGOON isn't having much success in the US, since there
is no marketing agency within the US.  If you know of a software house
that would liek to court TXT Inginerie to market DRAGOON within the US,
that would be great.  I think DRAGOON is a great language, and can grow
with more US usage.  Any takers?

dgw