Help me compare Ada-95 OOP to Java

[vsnyder@gyre.jpl.nasa.gov (Van Snyder)]

Some people have argued that the intellectual simplicity of Java's
"extensibility implies polymorphism" model compensates for a lack of
control over run-time dispatch.

How expensive is run-time dispatch?  That is, what's the incremental cost
of a call if the actual arguments are <type>'class instead of just <type>?

How much more difficult is it to master the difference between monomorphic
extensibility and polymorphic extensibility?

Object oriented facilities are presently under design for addition to
Fortran in the 2002 standard.  I'm arguing for a model as much like the
Ada-95 model as will fit in the confines of compatibility to Fortran-95.

Others appear to prefer something more like a hybrid of C++ and Ada-95:
Data are either monomorphic or polymorphic, and calls only dispatch if
polymorphic actual arguments are associated to monomorphic formals, but
the procedure call syntax looks like reference to a component of a record --
the Smalltalk "Send a message to the object" paradigm -- and the type of
only one formal argument of a type-bound procedure "extends" when the
procedure is inherited into a child type (the "receiver" argument).

Others want to go even further and adopt something more like the Java
model:  Extensibility implies polymorphism -- all objects of extensible or
extended type are polymorphic.

Do you have an opinion (or, better yet, measurements) of the benefits and
costs of these features in Ada-95?

E-mail would help me most.

-- 
What fraction of Americans believe   |  Van Snyder
Wrestling is real and NASA is fake?  |  vsnyder@math.jpl.nasa.gov