Re: Top 10 Ada myths

[dewar@cs.nyu.edu (Robert Dewar)]

Actually the real history of derived types is as follows:

In an attempt to simplify the Ada 83 definition, the DR's discussed the
possibility of eliminating derived types. Note that in this connection:

  type x is range 1 .. 10;

is not a derived type (in fact Ada 83 remains confused over whether this
is or is not a derived type declaration, although in Ada 95, it definitely
IS a derived type declaration).

At a meeting which Jean did not attend, the DR's voted unanimously (as
I remember with no abstentions) that derived types should be removed.

JDI was pretty horrified, since he felt that DT's were a basic building
block of the language, and vowed never to miss another DR meeting (Tuck
never missed a 9X DR meeting, so I guess he took this experience to heart
too :-)

At the next meeting, we again discussed derived types. I am operating from
slightly rusty memory now, but my memory was that the vote was N-2, with
the 2 being Jean and Lee McLaren, who I remember very well saying "I can't
think of any use right now of derived types, but I might well think of some
use later on, so I don't want to eliminate them now".

And there it stood, but nothing happened. Some months later I was in the
AJPO director's office (I am pretty sure it was Bill Carlson at the time, but
these memories do get jumbled). I asked him what happened to DT's. He said
that Robert Firth had found some fundamental objections to their removal.
That seemed strange, so we called Robert in England, and his reply was
roughly "No, I can't think of any fundamental objections, but I think
we should keep them in, for one thing they look like they will be lots of
fun to implement" or something like that (and ther definitely should be 
a smiley on that quote)!

Anyway they stayed, and basically they stayed because JDI insisted that
he felt they were critical, and he held the pen. There weren't many issues
on which battle lines were drawn quite so clearly, but this incident is
interesting because it sure puts the lie to the notion of design by
committee. 

And it is interesting that derived types have now become the fundamental
basis for adding object oriented stuff in Ada 95 :-)

For Ada 95, again, although we had a committee, or really two (the DR's and
ISO WG9), Tuck kept a very firm hand on the design. THat is not to say he
didn't listen to input, from the committees and from the world, this was
a *very* open process. It is also not to say that the committees had no
effect. 

Both the DR's and WG9 had a much more conservative view of what should be
added to Ada 83 than Tuck, and reacted horrified to the scope of some of 
the earlier suggestions (see for example mapping document version 2). And
there is no doubt that the language was scaled back in response. But this
was still by no means design by committee. On the contrary, the result of
these discussions was to send Tucker back to the drawing board to figure 
out how to retain the crucial shape of his design while cutting back the
scope in a consistent manner.

In retrospect, although it seemed a bit of an energetic tussle at the
time, I think the dynamics resulted in a very successful final design.
If Tuck hadn't pushed the envelope, I think the conservative inclinations
of the DR's would have resulted in something MUCH too timid (there were
quite a few people who felt that minor tweaks to Ada 83 would be sufficient).
On the other hand, I think if the committee had not pushed to keep things
simple, then we would have ended up with something much too complex.

Certainly no one could describe the process as design by committee! Both
Ada 83 and Ada 95 were designed by teams with captains, and as I once
said in a letter to Government Computer News, when Dave Parnas trotted
out the usual "designed by committee" complaint, any sports fan knows
the difference between a committee and a team!

Robert