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

[eachus@mitre-bedford.arpa (Robert I. Eachus)]

     There are two intertwined issues here.  Jean sees beauty as an
important facet of programming language design, and his major
disagreement with Tucker is on exactly this issue.  He feels that wide
acceptance of Ada 9X requires that it be more elegant than Ada 83.  At
the current state of the process this is clearly not the case, but
knowing the people on board, especially Mike Woodger, I have to
believe that the final standard document will be much more elegant
than the current version of the ILS.

      But elegance also applies to the programs written in Ada 9X.
Jean send a message to the list saying it was now time to consider the
style issues and "with null" was first on the list.  While this phrase
is now gone in some places if not all, several people including Jean
felt that the keyword should be class, not tagged, and should precede
not follow the "type XXX is."  (Ada 83 does have both: "task type FOO
is..." and "type BAR is limited private;" so consistancy is not an
issue.)

     At the WG9 meeting in Salem, there was a vote to decide between
the two proposals (by countries since this was an ISO meeting).  The
vote was two to two with six abstentions.  (The US delegation felt the
appropriate position was to abstain initially then throw our support
to the winning side.  Oops!)  I think that overall the vote really
reflected a feeling that there might be a better alternative, but that
there really was no major techincal reason to favor one over the
other. Unfortunately, at this point a better alternative would require
major technical advantages to make it in the language.

     Second, and the major reason I am writing this, please don't let
the presence mislead you about the OO features in Ada 9X.  Of course,
Jean feels that this misperception will kill Ada 9X...  The mechanisms
in Ada 9X provide the necessary missing key pieces in Ada 83 to
implement all OO idioms.  (With the possible exception of troublesome
uses of multiple inheritance.  At this point I am actually more
concerned that the MI mechanisms in Ada 9X will allow misuse, that
that any current MI design cannot be implemented.)

     Maybe I should elaborate.  In Ada 9X, it is possible/easy to
create classes intended as mix-ins.  Take it to an extreme, and all
but one root (class) type in your program can be implemented as a
mix-in.  That doesn't seem to lead to surprises.  It is also possible
to inherit from a single parent no matter how that parent was
designed.  Also a clear mental model.  When you mix the SI and MI
models, it is not clear that all the possible consequences are easily
predictable.  (But at least they are always well specified, which is
better than most languages with MI features.)

--

					Robert I. Eachus

with Standard_Disclaimer;
use  Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...