Well, on the user list they just hitting on one of the
deficiencies in the current verifier mechanism - namely, that we aren't
more specific than a message per level of the spec that was violated.
One thing I'd definitely like to add is some context, which could
perhaps be represented as $1 $2 parameters in the resource message or
something. Examples of using this could be the name of the offending
class or method. Or perhaps we could just add a colon on to the end of
the message, followed by any context information - I guess that would be
the easiest.
The most painful section in my mind is indeed the "valid RMI-IIOP
type" check, since it has to check dozens of things about each class,
recursively to all classes that are referenced. It's very difficult to
figure out exactly what caused the error.
BTW - do we prevent circular references causing nasty loops here?
Aaron
