I was thinking that with reifiable types, the class or object could get
a reference to a Class type that describes the type.  (I assume this is
what has to happen for reifiable types anyway.)

So for some class

public Point3D<class T implements double.class, float.class>
{
  T x, y, z;
}

the compiler will have to add a field for the class type.  This type
could be used to "treat primitives like objects."

It depends on exactly what operations one needs to perform on primitives
that are "like a class."  Determining those operations is probably the
hard part.

BTW, if Point3D is reifiable, what is the class type?  I don't think
it's Point3D anymore.  So, can I get it's type with a literal still?
Will something like "this.class" be available? Or will I have to call
"this.getClass()?"  It seems like the latter is also a requirement.
Plus, it might be an error to invoke a class literal on a class that
hasn't been parameterized yet.

Point3D.class == we don't know yet, throw an error.

This reifiable proposal for Java 7 has some funny corners.

(This one, I mean:

<http://tech.puredanger.com/java7/#reified> )

Yeah, that's true.  Bad choice of words on my part.

Well... it's talking about expressions there, and casts in particular.

  "Unlike C and C++, the Java programming language allows only certain
forms of expressions to be used as expression statements. Note that the
Java programming language does not allow a "cast to void"-void is not a
type-so the traditional C trick of writing an expression statement such as:

    (void) ... ;           // incorrect!"

By contrast, in method declarations, their "ResultType" can be void.  I
think outside of the language spec itself, "type" is just a handy word
here to refer to a concept which includes void.

<http://java.sun.com/docs/books/jls/third_edition/html/classes.html#8.4>

Void is a class. :)

  "The Void class is an uninstantiable placeholder class to hold a
reference to the Class object representing the Java keyword void."

Note they refer to void as a keyword.

I think however that referring to void as a type is not playing too fast
and loose with the language.  It's a type, it just can't be used in
expressions (the only one that can't).

"Type or void" might be a better way of expressing what I'm trying to
say here. So might "result type."   But they're also a tad clumsy in
English.  I think some latitude is needed, when not writing a spec, as
to how concepts are referred to, as long as the concepts are clear.

I may have crossed the line from clarity to obtuseness by referring to
void as a type without any other discussion how I'm using that term, but
I don't think the word "type" should never be applied to "void" because
of that.