Martin Carlyle wrote:

<<Believe it or not, even at the US Air Force Academy, someone
suggested just this week that we replace Ada with Java in our
curriculum.  I had a fit (moderate overstatement).  Nonetheless, I
told the Colonel (department head), "Whatever you think we need that
Java has, I will deliver it by next fall."  I think they were
surprised I could make such a bold claim.  In reality, I doubt that
I'll end up having to do any additional work at all.>>

I think that in an Ada versus Java battle for the hearts and minds of
academia, we who advocate Ada can be on the offense versus
the defense.  If someone suggests replacing Ada by Java, here are
some comments I'd make:

1) In Java, OOP is the only show in town, and if you want to write a simple
non-OO application you need to use some heavyweight concepts --
eg a noninstantiable class with a private constructor.  Ada supports
both OO and non-OO apps impartially, and writing a simple non-OO
app does not require advanced features.  Students should use a language
that is not biased towards a particular design style.

2) The Java thread model is inferior to Ada's tasking.  Race conditions
and deadlock are much likelier to occur in Java.  (E.g. Java suffers
from the "nested monitor" problem, whereas in Ada an implementation
that supports Ceiling_Locking avoids this issue.)  Despite the presence
of priorities in Java, the semantics of scheduling/priorities is
completely implementation dependent, whereas the Ada Real-Time
Annex defines scheduling semantics and pragmas for the user
to set various policies.

3) Among other omissions, Java lacks enumeration types, fixed-point types,
subtypes, strongly typed scalar types, a parameter passing mode that allows
updating a scalar parameter, named parameter associations, default
values for parameters, and (quite significantly) generics.

4) Java does not have general block structure.  Ada allows subprograms to
be declared directly in other subprograms.  In Java a method may not be
declared local to another method.  It may be declared in an inner class that
is local to a method, but that introduces some conceptual complexity.
I would think that from an instructional viewpoint the traditional Algol 60
style block structure would be the preferred mechanism to present.

5) Java does not allow a method to be passed as a parameter.  There are
workarounds but at least to me these have the look and feel of a workaround
and not a solution.

6) Java presents a skewed view of data structures, since it uses pointers
implicitly for all aggregate data.  IMHO a student should be exposed to
the basics of record types and array types and understand the issues
posed by a language that supports pointers explicitly.  Java has garbage
collection, which from an app delevoper's point of view is great but
from the viewpoint of teaching students about data structures is not good.
It's a little like deciding if a calculator should be allowed in an
elementary
school course that is supposed to teach arithmetic.  After the students
understand the concepts behind long division then they can use a calculator
to ease their work, but giving this to them before means that they might
fail to learn the basic concepts/issues.

7) Java has nothing close to Ada's support for interfacing to foreign code.
Java "native methods" are far weaker and less portable than Ada's
interfacing facilities.

8) Simple interactive IO (especially input) is surprisingly complicated.
If you thought that the Ada 83 need to instantiate a generic for simple
integer
IO was a pedagogical problem, look at the code in a Java application that
reads an integer value from the keyboard.

Of course it would be naive to suggest that Java has no technical
advantages.
Here are a few that Javaphiles may suggest, along with my comments:

1) Java has a lighter-weight syntax for OOP.  This is definitely true, and I
have
found that Java's approach to multiple inheritance through interface
implementation
is conceptually simpler than Ada's techniques (generics or access
discriminants).
But some of Java's succinctness comes from its merging a module spec and
a module body into one unit, the class.  Ada separates a spec and a body,
a better methodological approach.  Moreover, multiple inheritance
is not something that students need to do all the time.

2) Java has a more extensive class library than Ada.  Agreed (with some envy
:-)
although with several current Ada vendors (such as Intermetrics, Aonix and
GNAT)
one can interface to Java classes.

3) Java's JVM offers "Write Once, Run Everywhere" (or so says Sun :-)
There are some glaring exceptions to this rule in the Java language spec,
e.g. thread scheduling as mentioned above, and current implementations
of the AWT also show platform-specific behavior (not just different
look and feel).  Moreover, the Sun/Microsoft battle has users wondering
what the Java language actually will be.  In contrast, Ada is an ISO
standard,
and portability experience has been quite good.  And in any event the
portability benefit of the JVM can be achieved by generating JCode from
Ada source, as is currently supported by several Ada vendors.

4) Java is hot new technology, and universities need to teach it in order
to compete for students.  I understand this motivation, but my suggestion
would be to introduce Java _after_ students have learned Ada.  This way
they get to learn first a language that is design methodology-neutral
and that instills good habits of program development.

Ben Brosgol
Aonix
200 Wheeler Rd
Burlington, MA 01803
(781) 221-7317 (phone)
(781) 270-6882 (FAX)
[log in to unmask]

PS- With apologies for this mini-ad at the end of this message, and with my
SIGAda chair hat on, I want to remind fellow Team-Adians about the SIGAda
'98
Conference (formerly Tri-Ada) which will be held Nov 8-12 in Washington, DC.
Go to the URL http://www.acm.org/sigada/conf/sa98 for more information.
Articles, abstracts, panel and workshop proposals are due May 1, and
experience reports are due on June 5.