Issues for Java to IDL RTF Discussion Mailing List

List of issues (green=resolved, yellow=pending Board vote, red=unresolved)

List options: All ; Open Issues only; or Closed Issues only

Issue 1578: Public method declarations in ResponseHandler interface
Issue 1589: orb() method on InputStream
Issue 1590: Marshalling of Any, Context, Principal, and TypeCode
Issue 1591: Default Constructor for Stubs
Issue 1592: How to make IDLEntity types serializable
Issue 1593: Obtaining a default ORB
Issue 1594: SerializationInputStream and SerializationOutputStream
Issue 1595: Unicode example needs fixing
Issue 1596: Mapping for nested sequences --- fix example
Issue 1597: Need to emit ID pragmas after declarations
Issue 1598: Lexigraphic sort for field names is before mangling
Issue 1599: Narrow signature does not allow use for abstract interfaces
Issue 1600: Section 5.7.4 : mapped IDL issue
Issue 1601: Can toStub take a stub?
Issue 1602: Does toStub return most derived stub?
Issue 1603: Clarify rules for mapping server-side errors and exceptions
Issue 1604: The generated IDL in A2 needs to be changed
Issue 1606: mangling rule for name clashes in IDL needs to be defined
Issue 1607: Do we need to protect non-generated IDL files against multiple inclusion?
Issue 1608: Rules in section 7.4.4 for toStub IIOP/JRMP search not correct
Issue 1609: Should exportObject throw exceptions?
Issue 1610: PortableRemoteObject toStub should throw NoSuchObject Exception
Issue 1619: Abstract Interfaces issue
Issue 1620: Omission in section 5.4.5
Issue 1621: Stub methods for hashCode, equals and toString
Issue 1640: Should Portability APIs throw SystemException?
Issue 1641: Mapping of Java names that are IDL keywords
Issue 1642: Java to IDL identifier mapping
Issue 1662: read_AbstractObject and write_AbstractObject
Issue 1736: Local stubs proposal
Issue 1780: JTS exceptions
Issue 1891: Need to #include orb.id
Issue 1894: Value methods not mapped in spec examples
Issue 1931: Need overloaded write_Value method on OutputStream
Issue 1957: Mapping of non-public types
Issue 1958: Mapping of java.rmi remoteException superclasses
Issue 1960: Mapping of Java arrays
Issue 1965: Eliminating server wait code
Issue 2081: Specifying code downloading
Issue 2082: Mapping of RuntimeException
Issue 2088: The mapping of java.lang.Class to IDL is not specified
Issue 2089: Custom marshalling wire format
Issue 2090: Mapping of Object, Serializable and Externalizable
Issue 2091: Support for serialPersistentFields
Issue 2092: RMI/ORB marshalling interface
Issue 2111: mapping from java.rmi Remote to CORBA::Object
Issue 2225: javax.rmi package use
Issue 2370: Mangling of Java long type not specified
Issue 2466: Mapping private Java methods to IDL
Issue 2467: Change write_* names in Util class
Issue 2468: IIOP/JRMP stubs name clash
Issue 2469: IDLEntity exception types as parameters and results
Issue 2470: Completion status missing from exception string
Issue 2471: RMI/IDL Tie changes
Issue 2472: Interfaces and values should be mapped consistently
Issue 2473: Mappings for non-conforming types
Issue 2474: Mapping for non-conforming class
Issue 2475: Definition of Stub.equals method
Issue 2476: IDLEntity types need to be boxed
Issue 2477: RMI Repository ID proposed change
Issue 2478: Replaceability of javax.* APIs
Issue 2479: Use of "java" as top-level IDL module
Issue 2480: export/unexport errors not well specified
Issue 2481: mapSystemException should preserve original exception
Issue 2482: Mapping of java.rmi.Remote
Issue 2483: NotSerializableException for RMI-IIOP
Issue 2484: Need read_Value(clz) on InputStream
Issue 2503: Need getTarget method on Tie interface
Issue 2504: Misleading wording in section 28.5.1.4
Issue 2505: ::java::lang::Exception is illegal IDL
Issue 2535: Mapping remote interface types to RMI-style repository IDs
Issue 2545: mapping of java.lang.Exception
Issue 2552: Treatment of classes with writeReplace/readResolve methods
Issue 2565: Mapping of Java constructors to init is broken
Issue 2804: stub classes can violate licensing and package sealing
Issue 2805:
Issue 2806:
Issue 2807:
Issue 2808:
Issue 2809:
Issue 2810:
Issue 2813:
Issue 2815:
Issue 2816:
Issue 2818:
Issue 2894: Security problem in JavaToIDL mapping
Issue 3093: PortableRemoteObject.narrow(null)
Issue 3117: Name mangling scheme broken.
Issue 3118: Boolean attributes
Issue 3151: Stream format problem for custom marshalled RMI types
Issue 3152: Tie.deactivate() exception handling
Issue 3160: Definition of serialVersionUID field
Issue 3200: Container/member name clashes
Issue 3433: No factory for javax.rmi.ClassDesc
Issue 3590: loader.loadClass() or Class.forName()?
Issue 3641: Java to IDL mapping contradictive to CORBA/IIOP 2.3.1
Issue 3670: Exception mapping needs fixing
Issue 3754: Java reverse mapping local case cannot properly support Portable Intercepto
Issue 3798: ValueHandler Interface
Issue 3857: Descriptions of readAny and writeAny not precise enough
Issue 3858: Package for PortableRemoteObjectDelegate
Issue 3969: Mapping of CORBA any and TypeCode
Issue 4004: Mapping CORBA minor codes to Java
Issue 4058: Mapping of Java byte constants to IDL
Issue 4296: javax.rmi.CORBA.Util.isLocal RemoteException condition(s) not specified
Issue 4319: Behavior of Java writeUTF/readUTF
Issue 4429: Issue with using the local stub with for example an EJB application
Issue 4497: Mapping subclasses of custom-marshaled Java classes to IDL
Issue 4591: New issue: Behavior of writeByte, writeChar, writeBytes, writeChars
Issue 4656: Mapping of ExceptionDetailMessage service context
Issue 4690: There is a typo on page 1-46.
Issue 4698: mapSystemException server mapping
Issue 4795: RMI-IIOP interop bet. J2SE 1.3 and 1.4
Issue 5330: Stream encoding for omitted optional custom data
Issue 5332: Urgent issue: javax.rmi.CORBA.Stub serialVersionUID
Issue 5454: Java Codebase tagged component and POA
Issue 5723: Issue with ValueHandler.getRunTimeCodeBase and PRO.exportObject
Issue 5741: 1.2.3 RMI/IDL Remote Interfaces 2
Issue 5742: scheme for mapping names is pretty byzantine
Issue 5772: Name mapping rules for non-ASCII characters
Issue 5991: Request clarification on copyObject(s) semantics
Issue 6410: Custom marshalled abstract interfaces and Java/IDL
Issue 6411: Custom marshalled abstract interfaces and Java/IDL
Issue 6992: defaultWriteObject clarification
Issue 7217: Mapping CORBA activity exceptions to Java
Issue 10336: Section: 1.3, Page: 1-21

Issue 1578: Public method declarations in ResponseHandler interface (java2idl-rtf)

Summary: This item affects the Java-to-IDL spec, but is also of concern to the
 IDL-to-Java RTF because it relates to the new portability stub APIs.
 
 Section 7.2.2 of the Java to IDL mapping specifies a new interface
 ResponseHandler with methods createReply and createExceptionReply.
 These methods are declared as public.
 
 Section 9.4 of the Java Language Specification states that "Every
 method declaration in the body of an interface is implicitly public.
 It is permitted, but strongly discouraged as a matter of style, to
 redundantly specify the public modifier for interface methods."
 
 I propose that the public modifier be removed from these method
 declarations.
 

Summary: This item affects the Java-to-IDL spec, but is also of concern to the
 IDL-to-Java RTF because it relates to the new portability stub APIs.
 
 An orb() method is needed on org.omg.CORBA.portable.InputStream for
 SerializationInputStream to call to find which ORB to use for 
 lookup_value_factory during execution of read_Value.
 
 I propose adding an orb() method to org.omg.CORBA.portable.InputStream.
 

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the IDL-to-Java RTF because it affects the IDL-to-Java
 mapping.
 
 How do SerializationInputStream and SerializationOutputStream marshal
 Any, Context, Principal, and TypeCode objects?  Is the same mechanism
 used as for user-defined IDL types, i.e., look for the IDLEntity
 marker interface and call the implementation helper class"s write
 or read method?  If so, we need to clarify that vendor implementation
 of these types must implement IDLEntity and provide a Helper class.
 
 I propose that we state in the spec that vendor implementations	of 
 these types must implement IDLEntity and provide a Helper class.
 

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the IDL-to-Java RTF because it affects the IDL-to-Java
 mapping.
 
 When adding support for idlToJava generated Stubs to 
 "InputStream.read_Object(java.lang.Class clz)" we ran into a small 
 problem while writing the code.  Initially we used "clz.newInstance()"
 to create an instance of the Stub object.  The problem here is the 
 Stub does not have a default constructor, so newInstance throws an
 exception.  We added a default constructor to the stub and everything
 worked fine.  The question is: should we add a default constructor to
 the stubs generated by idlToJava or should we use reflection(cost??) to
 invoke the one constructor that is already a member of the stub class?
 
 I propose that we state in the spec that a default constructor is
 required on all generated Stubs.
 
 

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the IDL-to-Java RTF because it affects the IDL-to-Java
 mapping.
 
 The Java to IDL mapping spec states that Java types produced by
 the direct mapping should implement IDLEntity and be serializable.
 This could be accomplished by having IDLEntity extend 
 Java.io.Serializable, or by having the generated types extend
 both IDLEntity and java.io.Serializable.  The spec should
 specify which of these approaches is to be used.
 

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the IDL-to-Java RTF because it affects the IDL-to-Java
 mapping.
 
 There are two places in the Java to IDL mapping spec that imply
 the use of a default ORB.  One is the toStub method of the
 PortableRemoteObject class, and the other is the readObject method
 of the javax.rmi.CORBA.Stub class.  It is not clear how we can 
 obtain a suitable default ORB for use in these APIs without either
 excessive resource consumption (if we start a new ORB every time)
 or violating applet isolation boundaries (if we use the same ORB
 every time).
 

There is a problem with implementing the deserialization of stubs
and the PortableRemoteObject.toStub method.  Both of these require
an ORB object to which the returned stub object will be connected
via its delegate.  However, there is no well-specified way to obtain
a suitable implicit ORB object in CORBA.

In the case of stub deserialization, it is not possible to pass an 
explicit ORB object to the deserialization operation.  In the case
of toStub, we do not wish to add an explicit ORB object because this
breaks the "portable RMI" programming model in which application
code refers only to RMI types and not to CORBA types.

Proposal:

Deserialization of javax.rmi.CORBA.Stub creates an "unconnected" stub.
Before being used, it must be connected to an ORB.  This will happen
automatically if the unconnected stub is written to a GIOP marshaling
stream by being passed to javax.rmi.CORBA.writeRemoteObject or
javax.rmi.CORBA.writeAbstractObject.  However, in other cases, the stub
must be connected explicitly.  To enable this, the method
   void connect(ORB orb) throws RemoteException { ... }
is added to the javax.rmi.CORBA.Stub class, and the method
   static void connect(java.rmi.Remote unconnected, 
                       java.rmi.Remote connected)
               throws RemoteException { ... }
is added to the javax.rmi.PortableRemoteObject class.

The Stub.connect method throws RemoteException if called on a stub that
is already connected to an ORB (i.e., has a delegate set).  Otherwise,
a delegate is created for this stub and the specified ORB.  This method
is not intended to called directly by application code; instead,
application code should call the PortableRemoteObject.connect method,
which will in turn call the Stub.connect method.  This allows the
application code to remain portable between IIOP and JRMP.

The PortableRemoteObject.connect method may take an RMI/IDL stub or
an exported RMI/IDL implementation object as its "unconnected" and
"connected" parameters.  If "unconnected" is a connected object 
(i.e., an RMI/IDL CORBA stub with a delegate or an implementation
object with an RMI/IDL tie that has been associated with an ORB),
then a RemoteException is thrown.  Similarly if "connected" is an
unconnected object (i.e., an RMI/IDL CORBA stub without a delegate
or an implementation object whose RMI/IDL tie has not been associated
with an ORB), then a RemoteException is thrown.  Otherwise, "unconnected"
is connected to the same ORB as "connected" by setting its delegate if
it is a stub or by associating its tie with an ORB if it is an
implementation object.

RMI/IDL implementation objects may be connected implicitly by being
passed to javax.rmi.CORBA.writeRemoteObject or
javax.rmi.CORBA.writeAbstractObject, or explicitly by means of the
PortableRemoteObject.connect method.  Connecting an implementation
object is not the same as exporting it, and RMI/IDL implementation 
objects may be unconnected when first exported.  RMI/IDL implementation
objects are implicitly connected when they are exported to JRMP, and
RMI-JRMP stubs are implicitly connected when they are created.

The stub returned by PortableRemoteObject.toStub has the same
connection status as the target implementation object passed to toStub. 
So if the target object is connected, the returned stub is connected
to the same ORB.  If the target object is unconnected, the returned
stub is unconnected.

The javax.rmi.CORBA.Stub.writeObject method writes the following data
to the serialization stream:
 1. int - length of IOR type id  
 2. byte[] - IOR type ID encoded using ISO 8859-1
 3. int - number of IOR profiles
 4. For each IOR profile:
    a. int - profile tag
    b. int - length of profile data
    c. byte[] - profile data

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the Objects By Value RTF because it affects the Java
 ORB implementation of Objects By Value.
 
 The currently specified delegation relationship between 
 SerializationInputStream and SerializationOutputStream and their
 underlying ORB stream classes is inefficient and awkward.
 It is awkward because it requires the ORB implementation of the
 underlying streams to know about their "wrapper" serialization
 streams and to be careful to always pass out a reference to the
 wrapper instead of a "this" pointer.  It is inefficient because
 all items (even primitives) that are written to these streams 
 need to go through an extra delegation step through the wrapper 
 class as the underlying stream does not know whether it can
 safely bypass the wrapper method.
 

Summary:  The Unicode example given in the spec (5.2.4) is x\u22aBy
     javac diagnoses \u22aB as an invalid character in an identifier.
 
     A Java identifier may only consist of Unicode characters that denote a
     letter or a digit in a language. \u22aB (apparently) does not.
 
     Try \u03bC instead. According to 8.2.1 it"s a Greek mu. javac is happy
     with it.
 
     Proposed resolution: fix example
 

Summary:  mapping for nested sequences:
     - error in section 5.6 sequence<seq_Y>
     - no occurrence of >> so no need for white space
 
     Proposed resolution: fix example in section 5.6
 

 resolved, close issue

Summary: 10. Need to emit ID pragmas after declarations
 

Summary: Lexigraphic sort for field names is before mangling
 
     Proposed resolution: clarify section 5.5.6
 

Summary: 12. Narrow signature does not allow use for abstract interfaces
 
     Proposed resolution: return java.lang.Object
 

Summary: In section 5.7.4 the mapped IDL for Thrower includes:
        FruitbatException getLastException();
     It should be
        readonly attribute ::omega::FruitbatException lastException;
 

 In section 5.7.4, example IDL, change

Summary:  Can toStub take a stub?
 
     Proposed resolution: yes, in which case it"s a no-op.
 

Summary: 15. does toStub return most derived stub, or iterate looking for one?
 

 received issue

Summary: 17. Need to clarify rules for mapping server-side errors and runtime
     exceptions back to client exceptions.
 
     Proposed resolution: use omg.omg.CORBA.UNKNOWN
 
 

 issue resolved and fixed

Summary: The generated IDL in A2 needs to be changed.
 
     In particular, forward declares for fred.Test2 and fred.Stuff must
     appear before #includes, before sequence defs and before the module
     definition. Their #includes must appear at the end. In fact, just follow
     the file layout rules described at the beginning of the appendix. 
 
     If this is not done, there is a danger that circular references will
     lead to undeclared types when the IDL is compiled.
 
     Proposed resolution: fix example
 

Summary: Need to define a mangling rule for name clashes in IDL between 
     constant/method/attribute.
 

Summary: Do we need to protect non-generated IDL files against multiple
     inclusion?
 
     Proposed resolution: no protection
 
 

 No.--close issue

Summary: Rules in section 7.4.4 for toStub IIOP/JRMP search aren"t quite correct.
     If the object has been exported as IIOP (i.e., an IIOP Tie exists),
     no attempt will be made to create a JRMP stub even if there is no IIOP
     stub class.
 
     Proposed resolution: clarify spec
 
 

Summary: 25. Should exportObject throw exceptions?
 

 received issue

Summary:  PortableRemoteObject.toStub should throw NoSuchObjectException if object
     has not been exported.
 

 received issue

Summary: The current Java to IDL mapping spec does not fully address support
 for abstract interfaces.  There is support for these in the
 Portability Interfaces chapter, but not in the RMI/IDL Subset or
 IDL Mapping chapters. 
 

Summary: The text in section 5.4.5 has no mention of String constants.
 This seems to be an accidental omission.  I propose replacing
 the text in this section by the text in section 5.5.5, with
 "value type" changed to "interface".
 

Summary: RMI/JRMP stubs provide overridden implementations for the
 hashCode, equals, and toString methods of java.lang.Object.
 
 The stub hashCode method returns the same hash code for
 different stubs that refer to the same remote object.
 The stub equals method returns true when stubs that refer
 to the same remote object are compared.  These methods allow
 remote objects to be held in Java collections.
 
 The stub toString method returns a string containing information
 about the remote object, not the local stub.
 
 RMI/IIOP stubs should provide equivalent methods.  The easiest
 way to do this is to add implementations of these three methods
 to the javax.rmi.CORBA.Stub base class.
 
 I propose that the definition of javax.rmi.CORBA.Stub in
 section 7.2.5 be modified to add these methods.
 

Summary: This item arises from the Java-to-IDL Mapping RFP, but is also of 
 concern to the IDL-to-Java RTF because it affects the IDL-to-Java
 mapping.
 
 Three of the new portability APIs intorduced by the Java to IDL mapping
 are declared as throwing org.omg.CORBA.SystemException.  These are
 the _invoke methods of ObjectImpl and Delegate and the _invoke method
 of InvokeHandler.  I don"t think this exception hould be declared
 explicitly, since it is a subclass of java.lang.RuntimeException and so 
 is always implictly throwable whether it is declared or not.  Also, 
 other portability APIs can throw this exception even though they do not 
 declare it explicitly.  This inconsistency makes it very difficult for
 the reader of the spec to determine which of the portability APIs can
 throw CORBA system exceptions, and is likely to cause confusion.
 
 I propose that org.omg.CORBA.SystemException be removed from the throws
 clause of the above-mentioned APIs.
 

Summary: The Java to IDL mapping spec says that a Java name that is an IDL
 keyword is mapped to a mangled name consisting of the IDL keyword 
 followed by a trailing underscore.  Now that the Objects By Value
 RTF has adopted escaped identifiers with keading underscores, we
 should revise this mapping to use escaped identifiers.
 I propose that section 5.2.2 of the Java to IDL mapping spec be
 changed to state that Java names that collide with IDL keywords 
 are mapped to IDL by adding a leading underscore, so that typedef
 would be mapped to _typedef.
 

Summary: orbos/98-02-01 says:
 
 	We have provided name manglings to work around [the limitations],
 	but we recommend that OMG consider extending the definitions of
 	IDL identifiers so that a wider range of unicode characters can
 	be supported and so that case is significant in distinguishing
 	identifiers.
 
 I would suggest to remove this recommendation because it is pragmatically
 unimplementable for implementation languages that do not support
 extended character sets. If the OMG were to take this step, it would
 effectively alienate CORBA from all such implementation languages, which
 I believe would be detrimental to the success of CORBA.
 
 	Because overloaded methods are a popular feature in object-oriented
 	programming languages we recommend that OMG considers extending
 	IDL to allow overloaded methods.
 
 Again, I suggest to remove the recommendation because it is pragmatically
 unimplementable. Steve Vinoski recently posted an interesting article
 on this topic in comp.object.corba -- I have attached a copy below.
 

Summary: In section 7.1.2, there is a method called read_AbstractObject(clz).
 I propose that we change its name to read_Abstract to bring it
 into line with the read_Abstract() method introduced in section
 8.7.1 of the OBV spec.  This is consistent with having methods
 called read_Object() and read_Object(clz).
 
 There"s also a typo in section 7.2.6.  It says that 
 Util.write_AbstractObject calls OutputStream.write_AbstractObject.
 This should say OutputStream.write_Abstract.
 
 
 

Summary: There is currently no standard support for local RMI/IDL stubs.
 In contrast, the IDL/Java mapping specifies how support for local
 stubs is to be provided.  The Java to IDL mapping should also 
 specify standard APIs for local stubs which are compatible with
 the POA and the mechanisms used by IDL/Java local stubs.
 
 

Summary: The Java-to-IDL spec says in section 6.6.1 that certain CORBA system
 exceptions are mapped to the Java exceptions
   javax.jts.TransactionRequiredException
   javax.jts.TransactionRolledBackException
   javax.jts.InvalidTransactionException
 
 The names above are obsolete (since Sun revised the JTS/JTA specs),
 so the Java-to-IDL spec needs to be changed.  Possible mappings are:
  1. javax.transaction.<the_same_names_as_above>
       (as listed in the JTA spec)
  2. java.rmi.RemoteException
  3. No mapping (leave them as CORBA system exceptions) 
 
 Option 1 implies a dependency of the RMI-IIOP runtime on these
 JTA exception classes.  
 
 

Summary: The example in section A.2 of the Java-to-IDL spec uses the
 type ::CORBA::WStringValue but does not define this type.
 By CORBA convention, the definition will be in the file
 orb.idl, which should be #included in the example IDL code.
 I propose that this example be changed to #include orb.idl.
 l
 

Summary: The Java to IDL mapping spec specifies rules for how value
 methods are mapped to IDL.  However, none of the examples
 shows any of these methods actually being mapped.  Although
 this is legal according to the spec, it has led to some
 confusion as to which methods would normally be expected to
 be mapped, such as the writeObject method in section 5.5.9.
 
 

Summary: Suppose we are passing an array of declared type Animal[] and
 runtime type Wombat[] but whose elements are all of runtime type
 Wombat, where Wombat is a subtype of Animal.  This array is mapped
 to a boxed valuetype in the org.omg.sequences module.  In the
 RMI-IIOP stub, we need to write this array by making a write_Value
 call (so that sharing can be handled correctly).  The repository ID
 that we need to write is for a type seq_Animal, based on a static
 mapping from the declared type of the array (see section 5.6 of the
 Java-to-IDL spec).  However, if the stub used the normal write_Value
 call that takes a single argument of the object to be written, the
 runtime will not be able to put the correct repository ID on the
 wire because it only knows about the runtime type of the array, not
 the declared type.  For the IDL case, this is taken care of by 
 having the stub generate the form of write_Value that takes a
 ValueHelper object for the declared type, but in RMI-IIOP there are
 no ValueHelper objects.
 
 
 

Summary: The spec needs to say what happens if a Java public RMI/IDL
 value class has data members whose types are not public.
 
 Options: Either a) restrict all uses of non-public types, or 
 b) map all non-public types to public in IDL.
 
 Inprise have proposed that we adopt option b).  Comments?
 

Summary: There is an issue with the currently specified mapping for Java
 remote and abstract interfaces containing methods that throw 
 superclasses of java.rmi.RemoteException but do not throw
 java.rmi.RemoteException.  In Java 1.1, these are not recognized
 as RMI remote interfaces.  In Java 1.2, these are recognized as
 RMI remote interfaces. 
 
 I propose that throwing these superclass exceptions be treated by
 the Java to IDL mapping as semantically equivalent to throwing
 both RemoteException and the superclass exception.  This means
 that the section 4.3 rules for RMI/IDL remote interfaces would 
 be changed to include this case (in point 3), with similar changes
 elsewhere in the spec.
 

Summary: Generated names in the ::org::omg::sequences module can clash.
 For example, a Java array of type foo[][] and a Java array of type
 seq_foo[] both map to ::org::omg::sequences::seq_seq_foo.
 
 I propose that instead of using repeated "seq_" prefixes for the 
 array dimensions, a single prefix "seq<n>_", where <n> is the
 number of dimensions, should be used.  This ensures no clashes.
 In the example above, these Java types would map to seq2_foo and 
 seq1_seq_foo.
 

Summary: An RMI-JRMP server creates and exports server objects and then
 returns.  A non-daemon thread created by the JRMP runtime keeps the
 server process alive until there are no more objects exported.
 For code portability, an RMI-IIOP server needs to work in a similar
 way.  It is not acceptable to require a call to ORB.run() or the
 use of application "wait" code to keep the server process alive.
 
 I propose that the specification of PortableRemoteObject.exportObject
 be updated to say that the first call to this creates a non-daemon
 thread which keeps the JVM alive until all exported objects have
 been unexported by calls to PortableRemoteObject.unexportObject.
 

Summary: The spec should mandate codebase annotation when marshalling in Java,
 and should specify how the annotation is selected.
 
 The spec should mandate code downloading using the codebase annotation
 when demarshalling in Java, and should specify the classloader semantics.
 

Summary: If a Java method explictly declares that it throws RuntimeException or
 some subclass, the current spec says that these should be retained in
 the mapping to IDL.  I think this is wrong; they should be dropped
 when mapping to IDL.  The reasoning is along the same lines as dropping
 subclasses of RemoteException.
 

 issue resolved and closed

Summary: The mapping of java.lang.Class to IDL is not specified.  An
 interesting subquestion is whether it"s possible to transmit a
 Class that has not previously been mapped to IDL.
 

Summary: The full wire format produced by ObjectOutputStream really needs
 to be specified.  E.g., I would assume that anything written with
 writeObject gets marshalled as an Any.  In normal Java
 serialization, consecutive primitive writes (writeByte, writeInt,
 etc.) are aggregated into chunks with size lengths, and tagged
 to distinguish it from other data.  In this way, ObjectInputStream
 can ensure that major stream corruption never occurs: it can
 make sure that a primitive read never eats into object data,
 and vice versa.  Is a similar format used for IIOP?
 

Summary: The proposal here is to map Object, Serializable and Externalizable
 to distinct entities, rather than just to an any.
 
 These entities are
 module java {
     module lang {
        typedef any _Object;
     };
     module io {
         typedef any Serializable;
         typedef any Externalizable;
     };
 };
 
 and so
 java.lang.Object will map to ::java::lang::_Object,
 java.io.Serializable will map to ::java::io::Serializable, and
 java.io.Externalizable will map to ::java::io.Externalizable
 

Summary: Moving forward to JDK 1.2, the IDL value type data mapping needs
 to factor in the new serialPersistentFields declaration.  And there"s
 an interesting question whether the existence of a writeReplace method
 should influence the decision to map to a custom value.
 

 Adopted proposal:

Summary:  The ORB should be responsible for marshalling all value headers and
    other protocol specific data.  The RMI runtime should only marshal
    the state of the object.  In other words, we exepect the contract 
    between the ORB and RMI to be the similar to that of the ORB and
    an IDL generated ValueHelper.
 
  - The APIs should allow all RMI types supported by the IDL-Java mapping
    to be marshalled completely and correctly.
 
  - The APIs should be flexible enough to allow for some change in either
    RMI/Serialization semantics or changes to OBV wire protocol.
 

Summary: Are we trying to support sending JRMP stubs over IIOP?  If so, then
 it would seem that the IDL mapping of java.rmi.Remote to CORBA::Object
 works against that; it would need to be changed to Any.
 

Summary: It has concerned me for a while the use of the javax.rmi package in the Java
 to IDL mapping.  I think that PortableRemoteObject and other classes should be
 placed in a subpackage.  We may in the future put subpackages in javax.rmi and
 would like it organized much better than it is currently.  The
 PortableRemoteObject class should be put in a subpackage "portable" or
 something else.  I"d like to raise this as an issue to discuss at the RTF.
 There are still other issues that are being ironed out, and I think that this
 one can be addressed along with the others.
 

Summary: The overloaded method name mangling for the Java long type is not
 specified in the spec.  My suggestion is to use "long_long".
 

Summary: Java private methods and constructors should not be mapped to IDL.
 This has been agreed by the RTF, but an OMG issue number is needed.
 

Summary: In the javax.rmi.CORBA.Util class, the following names should be
 changed for consistency with the naming convention used in this package:
   write_RemoteObject to writeRemoteObject
   write_AbstractObject to writeAbstractObject
 This has been agreed by the RTF, but an OMG issue number is needed.
 

Summary: The names of generated stub classes can be the same in some cases for
 IIOP and JRMP.  This is a source of user error and confusion.
 

Summary: The specified Java to IDL mapping for IDLEntity exception types does
 not work for method parameters and results, because IDL exception
 types cannot be passed as method parameters and results.
 
 

Summary: The mapping of CORBA system exceptions to Java remote exceptions
 omits the completion status information.  This should be preserved
 in the detail string.
 

Summary: The following changes will improve the code generated for RMI/IDL Ties:
 
 a) RMI/IDL Tie classes shall catch org.omg.CORBA.SystemException and
    rethrow it (unwrapped by UnknownException).
 
 b) Change the signature of Util.mapSystemException to
      public static java.rmi.RemoteException 
                mapSystemException(org.omg.CORBA.SystemException ex);
 
 c) RMI/IDL Tie classes shall throw the result from mapSystemException.
 
 d) mapSystemException shall return the mapped exception if it is an
    instance of RemoteException or a subclass, and shall throw the
    mapped exception in all other cases.
 
 This has been agreed by the RTF, but an OMG issue number is needed.
 

Summary: The Java to IDL mapping rules for remote interfaces and value types 
 should be consistent except where there is clear need for differences.
 

Summary: The mapping for methods and constants in non-conforming types
 should be specified (by cross-reference to the words for value types).
 

Summary: The mapping of a non-conforming class should be abstract valuetype,
 not valuetype.  This is because instances of this type are not
 serializable, but instances of its subtypes are serializable.
 

Summary: The definition of the equals method of javax.rmi.CORBA.Stub should
 say that this method returns false when used to compare stubs that
 do not represent the same remote object.
 

Summary: When IDLEntity types are mapped from Java to IDL, they need to be
 mapped as valuetypes to preserve RMI null and sharing semantics.
 For most IDLEntity types, this will require the generation of
 boxed valuetypes in IDL, similarly to the mapping of Java arrays to
 IDL boxed sequences.
 

Summary: For the scoped name component, it is far simpler to use the actual class
 name of the java class with all the invalid IDL characters unicode escaped.
 

Summary: The javax.* APIs in the Java to IDL mapping spec have not been
 designed to separate interface from implementation.  For the RMI-IIOP
 standard extension, this is not a problem since the whole standard
 extension can be replaced by other vendors if necessary.  However,
 when RMI-IIOP becomes part of the core, it may be necessary to permit
 other vendors to replace these implementations without replacing the
 APIs (as is currently the case for the ORB APIs and implementation).
 

Summary: The Java to IDL mapping specifies a number of names in the IDL
 module "java".  Examples are ::java::lang::Exception and
 ::java::lang::_Object.  This can cause problems when mapping
 Java to IDL and then remapping the resulting IDL back to Java.
 

Summary: The text describing PortableRemoteObject does not describe the
 possible error cases very completely or precisely. 1. Change the paragraph:
 
    It is an error to call exportObject more than once for the same object.
 
    to the paragraph:
 
    It is an error to call exportObject on an object that is already
    exported.
 
 2. Change the sentence in the description of toStub:
   
    The argument object must either be a subclass of PortableRemoteObject
    or have been previously the target of a call on 
    PortableRemoteObject.exportObject.
 
    to the sentences:
 
    The argument object must currently be exported, either because it is
    a subclass of PortableRemoteObject or by virtue of a previous call 
    to PortableRemoteObject.exportObject.  If the object is not currently
    exported, a NoSuchObjectException is thrown.
 
 3. Change the sentence:
 
    The unexportObject method is used a deregister a server object from
    the ORB runtimes, allowing the object to become available for 
    garbage collection.  
 
    to the sentences:
 
    The unexportObject method is used a deregister a currently exported
    server object from the ORB runtimes, allowing the object to become 
    available for garbage collection.  If the object is not currently 
    exported, a NoSuchObjectException is thrown.  
 

Summary: The RemoteException returned by mapSystemException should preserve
 the original CORBA system exception as the detail field.
 

Summary: This issue was raised by Vijay Natarajan of Inprise.
 
 It appears that we need to do the same trick we used to distinguish
 between Anys, java.lang.Objects, Serializable etc
 
 We need to map java.rmi.Remote to be an alias of Object as in
 module java {
     module rmi {
         typedef Object Remote;
     };
 };
 
 This is to allow us to distinguish in IDL mapping between
 
 public interface hello implements java.rmi.Remote {
   void foo (java.rmi.Remote arg);
   void foo (org.omg.CORBA.Object arg);
 }
 
 w/o the aliasing, these methods cannot be mapped.
 

Summary: If a user passes a non-remote and non-serializable object to a remote
 method whose signature type is an abstract interface, the IBM/Sun 
 implementation of RMI-IIOP currently returns a ClassCastException.
 
 It is not very obvious to the user what is causing this error.
 RMI-JRMP returns a NotSerializableException in this case, which is
 much easier for the user to interpret.
 

Summary: It is possible for an RMI/IDL value class to contain an IDLEntity type
 that was mapped from an IDL valuetype.  In this case, the GIOP
 encoding of the IDLEntity type sent by an IDL/Java or C++ application
 to an RMI-IIOP appplication may not contain type information (if the 
 actual type is the same as the declared type).
 
 In order to allow the ValueHandler to demarshal these IDLEntity types,
 an additional API call such as read_Value(Class clz) in needed on the 
 org.omg.CORBA.portable.InputStream class.  This allows the ValueHandler
 to pass the expected class (which it knows) to the InputStream.
 
 For symmetry, a write_Value(java.io.Serializable obj, Class clz) method
 should be added to org.omg.CORBA.portable.OutputStream.  This would
 allow RMI-IIOP marshalling to use the same optimized encoding for this
 case as is possible for C++ and IDL/Java.
 

Summary: The javax.rmi.CORBA.Tie interface has a setTarget method but no getTarget.
 It is sometimes necessary to be able to obtain the target object for a tie.
 POA ties provide this facility, and it should be added to RMI/IDL ties.
 

Summary: The wording of section 28.5.1.4 is misleading.  It says that 
 writeRemoteObject and writeAbstractObject can allocate ties.  This
 is not incorrect, but it could be misinterpreted.  In some 
 implementations, a tie is allocated when an implementation object
 is exported, so this wording could be taken to imply that these
 methods do an "auto export" if an unexported object is passed in
 to these methods.
 

Summary: The ::java::lang::Exception type specified by the Java to IDL mapping
 is wrong.  It should be ::java::lang::_Exception, since exception is
 an IDL keyword.
 

Summary: RMI remote interface types are currently mapped to IDL interfaces without
 a #pragma ID, which means they have IDL-style repository IDs.  This causes
 problems when trying to locate a "best match" RMI stub for an IOR that
 contains this format of repository ID.  Because of the non-reversible
 manglings from Java interface names to IDL interface names (inner classes,
 leading underscores, and Unicode characters), there is no reliable 
 demangling from the IDL-style repid in an IOR to an RMI stub class name.
 
 This can be fixed by using the RMI style of repid instead of the IDL style
 of repid when mapping RMI remote interfaces to IDL, just as we currently
 do for value types.
 

Summary: I took another look at 28.3.7.1, and I think that the current
 special-case mapping of java.lang.Exception is odd and unnecessary.
 It seems to me that the right thing to do is to simply remove this
 special case, and map it the same as any other exception.  So,
 java.lang.Exception should map to a valuetype that has
 ::java::lang::Throwable as a superclass and no data fields, and to an
 IDL exception of ::java::lang::Ex (note that this exception is used in
 an example in 28.3.4.6).  java.lang.Throwable should map to a valuetype
 that has no superclass and a private data field named detailMessage, and
 to an IDL exception of ::java::lang::ThrowableEx.
 

Summary: We should consider whether a Java class that has a writeReplace or a
 readResolve method (with the appropriate signature for use by object
 serialization) should map to a custom valuetype in IDL instead of
 a normal valuetype.  We should also think harder about the impact
 of writeReplace/readResolve semantics on non-Java ORBs: can non-Java
 implementations properly receive and send instances of such classes,
 maintaining all the proper sharing, without running into type checking
 problems, and without any additions to non-Java ORBs?
 

Closed, no change. See revised text below.

Summary: The Java to IDL mapping specifies that Java constructors are
 mapped to IDL initializers.  Now that core RTF issue 1981 has
 replaced IDL initializers by factory methods, this mapping is
 broken and inconsistent with the core chapters.
 

Summary: Placing generated stub classes in the package in which the interface is defined
 is flawed for a few reasons:
 
 1) It can violate license agreements (e.g., Jini uses remote interfaces, and
    if an RMI-IIOP stub is generated from one of the core Jini interfaces 
    the license agreement will be violated since a new public class
    will be added to the core).
 
 2) Placing generated classes in the same package as the remote interface
    definition will not be possible if the package is sealed.
 
 If an interface is public, such generated classes could be placed in a
 sub-package of a CORBA-specific package, for example,
 org.omg.CORBA.stub.java.rmi._Remote_Stub.
 
 

Summary: We have a security hole in the reverse mapping that can be fixed by a
 straight forward spec change.
 
 The Java-to-IDL mapping spec 99-03-09 defines in paragraph 28.4.9.5 that
 rmi-iiop offer a method
 
 
 >     Util.loadClass(String className, String remoteCodebase, Class
 loadingContext)
 >         throws ClassNotFoundException { ... }
 >
 
 
 which does the following if everything else fails to load the desired
 class (on both JDK 1.1 and 1.2)
 
 
 > If a class was not successfully loaded by step 1, 2, or 3, and
 loadingContext
 > and loadingContext.getClassLoader() are both non-null, then call
 > loadingContext.getClassLoader().loadClass(className)
 
 
 By virtue of being a "public" core method as part of the rmi-iiop
 standard extension this mechanism essentially bypasses the "caller class
 loader" security check from Class.getClassLoader() by allowing arbitrary
 code to load a class using the class loader of any other class. The
 ClassLoader javadoc says to this effect.
 
 > Class loaders may typically be used by security managers to indicate
 > security domains.
 
 The fix for this situation is to change the signature of the above
 method to
 
 
 >     Util.loadClass(String className, String remoteCodebase,
 ClassLoader loader)
 >         throws ClassNotFoundException { ... }
 >
 
 so that the caller has to perform the necessary
 loadingContext.getClassLoader() call before calling this method. This
 way the callers security permissions are checked properly by the
 getClassLoader mechanism.
 
 So code that uses this API right now has to change from
 
 Util.loadClass(className, remoteCodebase, loadingContext)
 
 to
 
 Util.loadClass(className, remoteCodebase,
 loadingContext.getClassLoader())
 
 after this spec change.
 

The spec does not define what happens when a null object reference is passed to 
PortableRemoteObject.narrow().  For consistency with xxxHelper.narrow(), I think
it should return null.

Say I have a class octet in the global scope and I have an interface

interface foo extends java.rmi.Remote {
    void op(byte x);
    void op(octet y);
};

The operationsin IDL will now be

op__octet (octet x) and op__octet(::_octet);

which is incorrect.

This problem occurs in two cases only:
1. The case described in the issue, where two overloaded Java methods have parameters declared as byte (a Java primitive) and
octet (a class in the default package). The method name mangling rules map both of these to "octet".
2. The case where two overloaded Java methods have parameters declared as char (a Java primitive) and wchar (a class in the
default package). The method name mangling rules map both of these to "wchar".
 
There are many other cases of valid Java method declarations that do not map to unique names in IDL. The Java to IDL mapping
was never intended to ensure uniqueness in all possible cases, but only in cases that are likely to occur in real applications. Since
this error can only occur when an application contains classes called "octet" or "wchar" in the default package, and since use of
classes in the default package is not recommended for deployment of production applications, it is not worth introducing a new
mangling rule to support these two specific cases. Instead, the rules in section 1.3.2.9 should apply in this case.
(Vijay)
I vote no on 3117 and 3670 as I am very concerned that there are is an increasing number of cases where we decide not to handle
it because it's too difficult or backward incompatible and the number of cases where the reverse mapping cannot handle a valid
java class is becoming large.

Thought I would raise this as a formal issue.

The spec is ambiguous about mapping boolean accessors when in the following cases


boolean isFoo ();
void setFoo(boolean x);


and

boolean getBar();
boolean isBar();
void setBar(boolean x);

According to the spec, both of the above get mapped to attributes. But, an attribute has only one getter and one setter.
So, the question is what is isBar represented as on the wire? And how is the distinction made between isBar and getBar?

Two ideas I can think of are
1. Drop isBar from being recognized as a getter for attribute bar.
2. Have isBar be represented on the wire as _get_bar. This has the effect of the two methods collapsing into
one on the receiving context, which is probably OK because they should semantically be equivalent anyway.

There is a problem in the stream format for custom marshalled RMI types.
The Java serialization spec requires that when a class with a writeObject
method calls defaultWriteObject and then writes optional data, the receiver
must be able to skip the optional data if the readObject does not consume
it or the class is not in the receiver's hierarchy.  So if the sender's
hierarchy consists of serializable classes Base and Derived, both of which
call defaultWriteObject and then write optional data, there must be a way
for the unmarshaller to first skip to the end of the Base data, then skip
to the end of the Derived data.  With the current stream format, this is
not possible.

The javax.rmi.CORBA.Tie.deactivate() method does not declare any exceptions.
However, when using a POA servant, the following exceptions can be thrown by
the RMI/IDL tie's implementation of deactivate():

   org.omg.PortableServer.POAPackage.ServantNotActive
   org.omg.PortableServer.POAPackage.WrongPolicy
   (by the servant_to_id() method)

   org.omg.PortableServer.POAPackage.ObjectNotActive
   org.omg.PortableServer.POAPackage.WrongPolicy
   (by the deactivate_object() method)

How should these exceptions be handled by Tie.deactivate()?  If they should
be rethrown by Tie.deactivate(), what exception(s) should be used?

(This issue was raised by Max Mortazavi of Sun.)

The lack of a throws clause on the signature for Tie.deactivate() was an oversight. This method is called only by
Util.unexportObject(), which is called only by PortableRemoteObject.unexportObject(). Since the signature for
PortableRemoteObject.unexportObject() has a throws clause for java.rmi.NoSuchObjectException, the same throws clause should
be added to the signatures for both Tie.deactivate() and Util.unexportObject(). If a POA-based tie gets a POA exception when
deactivating an object, it should throw a NoSuchObjectException whose detail field contains the original POA exception.
Util.unexportObject() should rethrow this exception to PortableRemoteObject.unexportObject(), which should rethrow it to its
caller.

Does the serialVersionUID field need to be private, static, and final,
or just static and final?  The Java to IDL spec says private, static,
and final, but as far as I can tell, regular Java serialization just
uses static and final.  This has been a source of confusion for a number
of users.

There is a commonly occurring name collision that is not covered by the 
current name mangling rules.  This is demonstrated by the Java to IDL mapping
of java.util.Date.  The IDL generated gives a custom valuetype "Date" which 
contains an attribute "date", which is illegal in IDL.  This problem is not
unique to java.util.Date.

More generally, this collision occurs whenever a Java interface or class
being mapped to IDL contains a method or data member with the same name
as the interface or class (using a case-insensitive comparison).  For example,
the following Java definitions cause this problem:

There appears to be no way to construct a javax.rmi.ClassDesc, since the
ClassDesc class has two private members and no constructor. This is required by
non-default implementations of ValueHandler.

There are many ways to solve this issue.
1. Add a public constructor to javax.rmi.ClassDesc (preferred)
2. Make members protected so other implementations can subclass and add
constructors
3. create a factory method in say Util that returns a ClassDesc which takes the
required parameters.

It is not correct to say that non-default ValueHandlers must construct a ClassDesc instance, since the correct wire representation
can be written without an actual instance being constructed. However, it seems reasonable to modify the definition of the
ClassDesc class to allow instances to be constructed and initialized. The simplest way to do this is by making the data members
public instead of private. This also has the benefit of making these data members accessible from other languages, by making
them public in the Java to IDL mapping of this class. 
Note that changing these member declarations to public changes the default SUID. Therefore, for on-the-wire compatibility with
current implementations, it is necessary to specify the current SUID value using an explicit declaration.

In 1.4.9.5, #4 of the Java 2 case:

4. If a class was not successfully loaded by step 1, 2, or 3, and loader is
   non-null, then call loader.loadClass(className)

Shouldn't that really be Class.forName(className, false, loader), so that
array types are handled properly?

ClassLoader.loadClass does not handle creating new array types (it
will only find previously created array types), Class.forName does.

The Java to IDL Language Mapping (formal/99-07-59) seems to contradict the CORBA/IIOP 2.3.1 Specification (formal/99-10-07) over exceptions being raised by valuetype initialisers.

Section 1.3.5.4 of 99-07-59, point 3 states that exceptions thrown by a constructor of an object-by-value class are mapped to the corresponding OMG IDL exception.

On the other hand, section 2.8.1.5 of 99-10-07 does not allow initialisers to include a raises clause (contrast this with the syntax for an operation declaration in section 3.12).

This is a significant inter-operability problem. In my case the IDL generated by the IBM/SUN RMI-IIOP Java-to-IDL compiler (which includes 'raises' clauses in valuetype initialisers) is rejected as syntactically invalid by the MICO IDL-to-C++ compiler.
submit: Submit Issue Report

This issue could be resolved by adding exceptions to initializers in the CORBA core spec. However, the core RTF recently voted to close issue 3781 with no change because they considered adding exceptions to initializers to be out of scope for an RTF since it is not a bug fix. Also, it is very dificult to do this without causing incompatibilities in the IR interfaces in the CORBA module that deal with initializers. However, the Components RFP is adding an ExtendedIR module to accommodate the IR changes needed for CCM, and it is possible that the Java to IDL RTF could work with the Components FTF to add exceptions on initializers in the ExtendedIR module as part of CORBA 3.0 (though this change is not motivated by a components issue). One problem with this approach is that adding exceptions on initializers in CORBA 3.0 does not resolve the current broken state of the Java to IDL mapping in this area for CORBA 2.x. 
 
Before taking this issue any further with other RTFs or FTFs I would like to know if the Java to IDL RTF wants to pursue getting exceptions on initializers added to the core spec. Please indicate your preferred course of action by voting for either option A or option B below.
A. Replace numbered paragraphs 3 and 4 in section 1.3.5.4 of the Java to IDL mapping spec by the following text:
3. Java exceptions are not mapped to IDL.
B. Leave this issue open for now and work with other RTFs or FTFs to attempt to get the core spec changed to add exceptions to IDL initializers.

When I have Java exceptions called

foo
fooException
fooError

they all get mapped to the exception fooEx causing collisions.

I am not sure what the fix should be so I'll dump my thoughts here...

The correct fix for this probably should be:

foo -> fooEx
fooException -> fooExcp
fooError -> fooEr

However, given that that would not be backward compatible, another solution
would be

All exceptions map to the Ex suffix. Each collision adds an _ (underscore) at
the end, if there is a collision.

However, this has problem with the name changing depending on which exception
gets processed first, and the resultant definitions are not deterministic.

Any other ideas?

There is no conflict between fooError and the exception class names foo and fooException, since a trailing Error string is not
removed before appending the Ex suffix. However, there is a conflict between foo and fooException. 
Without making an incompatible change, there is no way to resolve this conflict, since the mapping of each class must be
deterministic and cannot vary depending on whether another exception class with this name conflict also exists (since the second
class could be created after the first class has been mapped). 
This is similar to the treatment of Java class or interface names that differ only in case. The mapping deals with such method or
field names within a class by applying a deterministic mangling rule, since all method and field names within a class are known
when the class is mapped. However, applying the same mangling to class or interface names would require global knowledge of
all class and interface names that could be mapped, which is not practical. 
It seems reasonable to follow this precedent and not specify a mangling fpr conflicting exception names that would require global
knowledge of all class and interface names that could be mapped. 
In practice this conflict is unlikely to cause any serious problems, since the JDK names all exceptions with an Exception suffix and
other packages should use a a consistent naming convention for all their exceptions (either with or without an Exception suffix). 
(Vijay)
I vote no on 3117 and 3670 as I am very concerned that there are is an increasing number of cases where we decide not to handle
it because it's too difficult or backward incompatible and the number of cases where the reverse mapping cannot handle a valid
java class is becoming large.

The local mapping makes it impossible to determine which ending point
to call.

ptc/00-01-06 says:

BEGIN_QUOTE

1.5.2.2 Local Stubs

The stub class may provide an optimized call path for local server
implementation objects. For a method echo(int x) of a remote interface
Aardvark, the optimized path does the following:

1. Find out if the servant is local by calling Util.isLocal()
2. If the servant is local, call 
	this._servant_preinvoke("echo",Aardvark.class)
3. If _servant_preinvoke returned a non-null ServantObject so, call
	((Aardvark)so.servant).echo(x)
4. If _servant_preinvoke returned a non-null ServantObject so, call
	this._servant_postinvoke(so)
5. If _servant_preinvoke returned null, repeat step 1. The call to
	Util.isLocal() will return false, causing the non-optimized
	path to be followed.
...

The following is an example of a stub class that provides this
optimized call path.

// Java
public class _Aardvark_Stub extends javax.rmi.CORBA.Stub
	implements Aardvark 
{
    public int echo(int x)
		throws java.rmi.RemoteException, Boomerang 
	{
	if (!javax.rmi.CORBA.Util.isLocal(this)) {
		...
	} else {
		// local call path
		org.omg.CORBA.portable.ServantObject so =
			_servant_preinvoke("echo", Aardvark.class);
		if (so == null)
			return echo(x);
		try {
			return ((Aardvark)so.servant).echo(x);
		} catch (Throwable ex) {
			Throwable ex2 = (Throwable)
				javax.rmi.CORBA.Util.copyObject(ex, _orb());
			if (ex2 instanceof Boomerang)
				throw (Boomerang)ex2;
			else
				throw javax.CORBA.Util.wrapException(ex2);
		} finally {
			_servant_postinvoke(so);
		}
	}
    }
}

END_QUOTE


ClientRequestInterceptor::send_request would need to be invoked by the
ORB inside _servant_preinvoke.

ClientRequestInterceptor::receive_reply or receive_exception would
need to be called inside _servant_postinvoke.

(Note that receive_other would not happen since, if a POA were
involved inside _servant_preinvoke and a servant manager raised
ForwardRequest, this should cause _servant_preinvoke to return null
resulting in a recursive call to the method.  In this case the ORB
would not call send_request in the first place, or, if it did, it
would handle the call to the receive_other internally inside of
_servant_preinvoke.)

One is tempted to say that the ORB could remember if
javax.rmi.CORBA.Util.copyObject was called on behalf of this request.
In that case, when _servant_postinvoke is called it would know there
was an exception (and would have it hands on the exception object -
which needs to be available via Portable Interceptors).

However, this does not work. The example does not show it, but if the
method returns values those values are filtered through
javax.rmi.CORBA.Util.copyObject before being returned to the client.
Exception objects are legal return values.  So there is no way to
determine if copyObject was called to copy return values or exception
objects resulting from an actually exception.  Therefore this trick
will not work.

Bottom line: there is no reliable way to determine which Portable
Interceptor ending point to call in the Java reverse mapping local
case. 

This issue is clearly about the Java to IDL mapping, so it belongs in the Java to IDL RTF, which is where is was originally. However, it was reassigned to this RTF I think largely because the solution requires changes to the org.omg.CORBA.portable APIs, which belong to the Java RTF. I have created issue 4701 to cover the Java mapping issues and the changes to the portable APIs. This issue should be re-assigned back to the Java to IDL RTF.

  In ptc/00-01-06.pdf,  the ValueHandler Interface shows method
implementations.  According to Java conventions, the methods
should not be implemented in interfaces.  It is requested that
this interface be updated to remove the method implementations.

The descriptions of readAny and writeAny in section 1.5.1.4 of the Java to IDL
mapping spec are not very precise and don't say how null values should be handled.
These methods are currently described in section 1.5.1.4 as follows:

> The writeAny method writes the Java object obj to the output stream out in the
> form of a GIOP any. The readAny method reads a GIOP any from the input stream
> in and unmarshals it as a Java object, which is returned.

Proposed resolution:

Replace the above paragraph in section 1.5.1.4 by the following:

 The writeAny method writes the Java object obj to the output stream out in the
 form of a GIOP any.  The contents of the GIOP any are determined by applying
 the Java to IDL mapping rules to the actual runtime type of obj.  If obj is null,
 then it is written as follows: the TypeCode is tk_value, the repository ID is
 "IDL:omg.org/CORBA/ValueBase:1.0", the name string is "", the ValueModifier is
 VM_NONE, the concrete base is tk_null, the member count is 0, and the any's value
 is a null valuetype (0x00000000).

 The readAny method reads a GIOP any from the input stream in and unmarshals it
 as a Java object, which is returned.  The following TypeCodes are valid for
 the GIOP any: tk_value, tk_value_box, and tk_objref.  If the TypeCode is anything
 other than these, a MARSHAL exception is thrown.

Using tk_value for java.lang.Object null seems incorrect because a java.lang.Object does not only contain valuetypes. Another
alternative is tk_null, but this is currently used for an any with no contents rather than an any containing a null. 
The bext compromise appears to be tk_abstract_interface. The IDL abstract interface type can represent all the legal values of
java.lang.Object in the Java to IDL mapping. Also, using this representation is consistent with the folllowing usability goals: 
1. When nulls are sent from IDL to RMI-IIOP, the same insertion operations for anys that are used for non-null values should
work for null values as well. This means that all the following anys when received by RMI-IIOP as the value of a java.lang.Object
should be unmarshalled as null: 

  tk_value+null

  tk_value_box+null

  tk_objref+nil

  tk_abstract_interface+null




2. When nulls are sent from RMI-IIOP to IDL, the encoding of null sent by RMI-IIOP should be extractable using the same
extraction operations that would be used for non-null values. For example, in the IDL-to-Java mapping, both the following should
work on null values sent by RMI-IIOP: 

  myValue = myAny.extract_Value();

  myObjRef = myAny.extract_Object();




So, if we say that tk_abstract_interface is the correct RMI-IIOP encoding for a null, then both extract_Value() and
extract_Object() should be able to handle a null tk_abstract_interface type in the any. The same should be true for other
languages. 
 
Note 1: it is not clear from either the Java language binding spec or the C++ language binding spec whether a value (not just a null
value) that was inserted using a tk_abstract_interface TypeCode can be extracted using the normal extract operations for
valuetype or object reference. If this does work (and there is good reason to think it should), then this would automatically take
care of handling a null encoded as tk_abstract_interface. 
 
Note 2: Issues like the one described in Note 1 above should not be up to all the language mapping RTFs to resolve. The core
should define semantics for anys that the language bindings are required to support. 
 
Note: (For information only) I have raised a separate issue in the IDL-to-Java RTF proposing that a value (including a null value)
that was inserted into an any using a tk_abstract_interface TypeCode should be extractable using the normal extract operations
for valuetype or object reference. If this passes, I will raise a similar issue in the C++ RTF

The Java to IDL mapping spec does not explictly specify the package for the
PortableRemoteObjectDelegate interface.

The correct fully-qualified name of this interface is 
javax.rmi.CORBA.PotableRemoteObjectDelegate.  This is stated explictly in
Harold Carr's final proposal for issue 2478, which contained the text:

> The implementation delegate classes for PortableRemoteObject and Util must 
> implement the following new interfaces: 
>
>       javax.rmi.CORBA.PortableRemoteObjectDelegate (per-class delegation) 
>       javax.rmi.CORBA.UtilDelegate (per-class delegation) 

and is also implied by the text that was voted on for issue 2478, which 
contained the following:

> The implementation delegate classes for PortableRemoteObject and Util must
> implement the following new interfaces for per-class delegation: 
> 
> package javax.rmi.CORBA;
> 
> public interface UtilDelegate {
> 
>     ....
> 
> }
> 
> public interface PortableRemoteObjectDelegate {
> 
>     ....
> 
> }

The above code clearly makes both interfaces part of the javax.rmi.CORBA package.
However, when these definitions were copied into the Java to IDL mapping spec,
the interfaces UtilDelegate and PortableRemoteObject were placed in separate 
sections 1.5.3.2 and 1.5.3.3, and the package statement was copied to section 
1.5.3.2 but was unfortunately not copied to 1.5.3.3.  This creates an ambiguity
in the published spec document, since section 1.5.3.3 does not explicitly specify
the package for PortableRemoteObjectDelegate.

Proposed resolution:

Add the line
   package javax.rmi.CORBA;
to section 1.5.3.3.

The Java to IDL mapping spec does not explictly specify the package for the PortableRemoteObjectDelegate interface. 
The correct fully-qualified name of this interface is javax.rmi.CORBA.PotableRemoteObjectDelegate. This is stated explictly in
Harold Carr's final proposal for issue 2478, which contained the text: 

> The implementation delegate classes for PortableRemoteObject and Util must 

> implement the following new interfaces: 

>

>       javax.rmi.CORBA.PortableRemoteObjectDelegate (per-class delegation) 

>       javax.rmi.CORBA.UtilDelegate (per-class delegation) 




and is also implied by the text that was voted on for issue 2478, which contained the following: 

> The implementation delegate classes for PortableRemoteObject and Util must

> implement the following new interfaces for per-class delegation: 

> 

> package javax.rmi.CORBA;

> 

> public interface UtilDelegate {

> 

>     ....

> 

> }

> 

> public interface PortableRemoteObjectDelegate {

> 

>     ....

> 

> }




The above code clearly makes both interfaces part of the javax.rmi.CORBA package. However, when these definitions were
copied into the Java to IDL mapping spec, the interfaces UtilDelegate and PortableRemoteObject were placed in separate
sections 1.5.3.2 and 1.5.3.3, and the package statement was copied to section 1.5.3.2 but was unfortunately not copied to 1.5.3.3.
This creates an ambiguity in the published spec document, since section 1.5.3.3 does not explicitly specify the package for
PortableRemoteObjectDelegate.

The Java to IDL mapping does not specify how the Java classes produced
by the forward mapping from the CORBA "any" and "TypeCode" types should
be mapped back to IDL.

Since org.omg.CORBA.Any and org.omg.CORBA.TypeCode both inherit from 
IDLEntity, section 1.3.9 implies that these types should be "boxed" in
the same way as other IDLEntity types.  This would imply the following
mapping to IDL:

module org {
module omg {
module boxedIDL {
module CORBA {
valuetype _Any any;
#pragma ID Any “RMI:org.omg.CORBA.Any:xxxxxxxxxxxxxxxx:yyyyyyyyyyyyyyyy”
};
};
};
};

module org {
module omg {
module boxedIDL {
module CORBA {
valuetype TypeCode ::CORBA::TypeCode;
#pragma ID TypeCode “RMI:org.omg.CORBA.TypeCode:xxxxxxxxxxxxxxxx:yyyyyyyyyyyyyyyy”
};
};
};
};

The above raises a number of questions:

1. What is the correct IDL name for the boxed valuetype enclosing the Java "Any"
   type?  Is it ::org::omg::boxedIDL::CORBA::_Any or ::org::omg::boxedIDL::_any?

2. How should the hashcodes and SUIDs be computed in the above repid pragmas?
   For other boxed IDLEntity Java types, these are computed from the actual
   implementation classes.  However, for these two types, the actual
   implementation classes are ORB-specific and it seems meaningless to send
   their hashcodes or SUIDs on the wire.  We could either do what we do for
   object references and send zeros for the hashcode and SUID, or we could
   use IDL-style repids instead or RMI-style repids for these types.

3. Is it better to make these types an exception to the general rule as defined
   in section 1.3.9?  An alternative non-boxed mapping would be to simply map
   them to the IDL/PIDL types "any" and "::CORBA::TypeCode".  This would have
   the drawback that RMI semantics of being able to pass nulls and maintaining
   referential integrity for shared references to "by value" Java objects would
   not apply to these Java types.

This resolution is being voted a second time with a minor change. The original resolution (which passed) named the mapped IDL
type as ::org::omg::boxedIDL::CORBA::_any. The amended resolution names the type as ::org::omg::boxedIDL::CORBA::_Any.
This is the only change to the original resolution. 
 
The following points are in favor of mapping org.omg.CORBA.Any and org.omg.CORBA.TypeCode to boxed IDL types: 
1. It is more consistent with the current spec, which says that all IDLEntity types are boxed, except for certain specified cases. 
2. It supports sharing semantics when passing multiple references to the same Java Any or TypeCode object across RMI-IIOP,
just like other IDLEntity and serializable objects. 

If these are mapped to boxed types, then we need to specify the repids for the boxed valuetype encoding. These would not be
derived from the Java implementation classes as they are for other IDLEntity types (e.g., RMI:com.acme.AnyImpl:hhh:sss) since
using implementation names would not be interoperable. It would seem reasonable to use 
  RMI:org.omg.CORBA.Any:0000000000000000 and 
  RMI:org.omg.CORBA.TypeCode:0000000000000000 
as the repids.

Section 1.4.8 of the Java to IDL mapping spec says that when mapping
CORBA system exceptions to RMI exceptions, the minor code is formatted
as a decimal number.  This is a readability issue for both OMG standard
minor codes and vendor-specific minor codes.  For example, the OMG
standard minor code 0x4f4d0001 would be formatted as 1330446337 and a 
vendor proprietary minor code 0xc9c25401 would be formatted as 3384955905
(or -910011391 if using signed 32-bit arithmetic). 

Proposed resolution:

In section 1.4.8, change the fourth bullet to:
 • followed by the hexadecimal value of the system exception’s minor code

In the following paragraph, change the example detail string to:
 “CORBA UNKNOWN 0x31 Maybe”

Since Java byte values are mapped bit-for-bit to IDL (see section 1.3.3),
the same should apply to byte constants.  For example, -1 should map to 255. 
However, section 1.3.4.5 says that these constants are mapped to the
same values.

Proposed resolution:

In section 1.3.4.5, change "are mapped to the same values" to "are mapped
to the same values, except for byte constants which are mapped bit-for-bit.
For example, -1 maps to 255."

The Java to IDL spec 
(http://cgi.omg.org/cgi-bin/doc?ptc/00-01-06, section 1.5.1.4 page 47)
says:

	"The isLocal method has the same semantics as the ObjectImpl._is_local
	method, except that it can throw a RemoteException."

However, it does not specify under what conditions a remote exception
is thrown.

Either the exception condition(s) should be specified or the exception
should be removed from the signature.

Java Serializable classes can define their own readObject/writeObject
methods which take java.io.ObjectInputStream and
java.io.ObjectOutputStream parameters, respectively.

What is the proper behavior of the readUTF and writeUTF methods on these
streams in RMI-IIOP?  The Java to IDL specification doesn't mention
these methods.

Proposed resolution:

In RMI-IIOP, java.io.ObjectOutputStream's writeUTF method should write a
CORBA wstring.  Similarly, readUTF should read a CORBA wstring.

Discussion:

One must override these methods to avoid using the internal java.io
implementations.  Looking at the JDK 1.3 code, the behavior is specific
to Java serialization.  If one doesn't override writeUTF and readUTF,
Java serialization mechanisms such as blocks (similar to chunks) may be
introduced on the wire in RMI-IIOP.

When using the local stub with for example an EJB application, we discovered the following critical issue. 

Let's take an simple example : 

public interface HelloWorld extends javax.rmi.EJBObject { void print( String message ); } 

The generated Stub according to the mapping rules contains the methods for the inherited interface as for example 'getEJBHome'. 

For a local invocation on 'getEJBHome' the following part of the stub will be used : 

public javax.ejb.EJBHome getEJBHome() throws java.rmi.RemoteException { while( true ) { if (!javax.rmi.CORBA.Util.isLocal(this) ) { // non local use } else { org.omg.CORBA.portable.ServantObject _so = _servant_preinvoke("EJBHome", HelloWorld.class); 

if ( _so == null ) return getEJBHome(); 

try {return ((javax.ejb.EJBObject)_so.servant).getEJBHome(); } catch ( Throwable ex ) { Throwable ex2 = ( Throwable ) javax.rmi.CORBA.Util.copyObject(ex, _orb()); 

throw javax.rmi.CORBA.Util.wrapException(ex2); } finally {_servant_postinvoke(_so); } } } } 

The '_servant_preinvoke' method will return a servant class of type 'HelloWorld' that implements the 'getEJBHome' operation. 

Thus, the '_HelloWorld_Tie' servant is returned. 

Then, the cast to convert the returned servant to 'javax.ejb.EJBObject' will fail since the servant doesn't implement this interface ( it only inherits from org.omg.PortableServer.Servant and implements javax.rmi.CORBA.Tie ). 

Thus, by using the local stub, it is not possible to invoke an inherited operation. 

Solution : 

The returned servant is always implementing 'javax.rmi.CORBA.Tie'. To allow the correct cast to 'javax.ejb.EJBObject' we need to access the final target instead of the servant itself. So the following solution can be always applied : 

javax.rmi.CORBA.Tie tie = ((javax.rmi.CORBA.Tie)_so.servant); 

return ((javax.ejb.EJBObject)tie.getTarget()).getEJBHome(); 

The Java to IDL mapping spec says that serializable Java classes that
have a writeObject method are mapped to IDL custom valuetypes.  This
should also apply to any Java subclasses of these classes, since otherwise
the generated IDL is illegal.  (An IDL non-custom valuetype can't
inherit from an IDL custom valuetype.)

Proposed resolution:

In section 1.3.5.6 of the Java to IDL mapping spec, change the first sentence
of the second paragraph from

If the class does not implement java.io.Externalizable but does have a 
writeObject method, ...

to:

If the class does not implement java.io.Externalizable but does have a 
writeObject method, or extends such a class directly or indirectly, ...

.

A Serializable or Externalizable can define methods which take
java.io.ObjectOutputStreams.  There are four methods on that class
that are well defined in Java, but the mapping to CORBA may need to be
clarified:

writeByte(int)
writeChar(int)
writeBytes(String)
writeChars(String)

Please see the Java docs for these methods:

http://java.sun.com/j2se/1.3/docs/api/java/io/DataOutput.html#writeByte(int)
http://java.sun.com/j2se/1.3/docs/api/java/io/DataOutput.html#writeChar(int)
http://java.sun.com/j2se/1.3/docs/api/java/io/DataOutput.html#writeBytes(java.lang.String)
http://java.sun.com/j2se/1.3/docs/api/java/io/DataOutput.html#writeChars(java.lang.String)

Based on those detailed definitions, I'd say the mapping would be:

writeByte(int)
* write_octet of the lower 8 bits of the int

writeChar(int) 
* two write_octet calls in the order as shown in the java docs

writeBytes(String) 
* For each char in the String, call charAt, take the lower 8 bits, and
call write_octet

writeChars(String)
* For each char in the String, call charAt, split the char into two
bytes, and make two write_octet calls in the order shown in the java
docs

Another interpretation might be to use wstrings or wchar arrays for
writeChars.  The problem I see there is that since there isn't a
readChars method, the user will be using the DataInput methods
readFully or multiple readChar calls to reconstruct the String.  Since
the wire format of wstrings/wchars depends on the code set, he might
start to see code set specific bytes instead of the expected bytes
detailed in the java docs of writeChars and writeChar.

As proposed in the issue summary, the stream data written by these APIs for RMI-IIOP needs to be consistent with Java serialization, since otherwise some classes that serialize and deserialize correctly using Java serialization or RMI-JRMP would not serialize and deserialize correctly using RMI-IIOP. 

However, care needs to be taken when making this change to avoid breaking existing applications. This is because 
the revised specification for writeBytes(String) is incompatible with the way this method is currently implemented by certain JDK ORBs (specifically IBM's J2SE 1.3 and Sun's J2SE 1.3), and 
the change may affect application code as well as the ORB. 
The incompatibility is that the writeBytes() method currently converts characters to bytes using the platform's default character encoding, but the new specification doesn't do this. Therefore, any platform whose default character encoding is not ISO8859-1 (e.g., IBM zOS) will send different data for writeBytes() with the revised specification, and any application-defined Java readObject() methods that consume this data (on any platform) will have to be modified to consume unconverted data instead of converted data. Similarly, IDL custom valuetype unmarshal method implementations that consume data written by writeBytes() may have to be modified in the same way. 

Following established principles for incompatible changes to the J2SE platform, this change should only be made with adequate warning and on a major release boundary of the J2SE platform (e.g, 1.4 or 1.5) rather than on a minor release boundary (e.g., 1.4.x) or a service update (e.g., 1.4.0_xx).

The interop RTF resolved issue 3405 by adding a new service context called
ExceptionDetailMessage.  The usage of this service context was left to
language mappings to define.  This usage has been defined for the IDL to Java
mapping (issue 4013) but not for the Java to IDL mapping.

The resolution for issue 4013 defines how this service context shall be used
when marshaling and unmarshaling system exceptions.  However, an RMI-IIOP client
never receives a system exception, but instead receives a mapped RemoteException
or RuntimeException that was created by applying the rules specified in 
section 1.4.8.  Currently these rules do not take account of any detail message
that the system exception has acquired from the ExceptionDetailMessage service
context by applying the mappings specified by the resolution of issue 4013.

In order to ensure that this valuable diagnostic information is available on
exceptions received by RMI-IIOP clients as well as IDL clients, I propose the
following change to section 1.4.8 of the Java to IDL mapping spec.

Proposed resolution:

After table 1-2, replace

  "In all cases, ..."

by

  "If the getMessage() method of the CORBA system exception returns a non-null
  and non-empty string, the RMI exception is created with this string as its
  detail string.  In all other cases, ..."

Resolution: 
Close, no change. The proposed change would cause more problems than it solves. This is because the getMessage() method of RemoteException returns its own detail message concatenated with the detail message of its nested exception. So copying the nested exception's detail message into the top-level RemoteException will cause the same message to be printed twice, which is not very useful!

"The mapSystemException method maps a CORBA system exception to a java.rmi.RemoteException or a java.rmi.RuntimeException." 

There is no java.rmi.RuntimeException, instead it should read: java.lang.RuntimeException !

.

Ie RemoteException and subclasses are not RMI/IDL Exception types. (Note
> > >however,
> > >that the section refered to doesn't explicitly state that. That is a spec
> > >problem). The spec provides these options:
> > >If the CORBA system exception org.omg.CORBA.portable.UnknownException
> > >is thrown, then the stub does one of the following:
> > >- Translates it to org.omg.CORBA.UNKNOWN.
> > >- Translates it to the nested exception that the UnknownException contains.
> > >- Passes it on directly to the user.
> > >
> > >This option is needed because, some userdefined runtime exception, say,
> > >may not
> > >have an equivalent implementation on the receiving end. If we fix remote
> > >exception to always result in the "right" exception on the wire, this case
> > >occurs
> > >only when there is an unexpected exception (such as a null pointer), which
> > >should
> > >be OK. The intent here was to talk about base classes and not individual types, but I
> > >agree this is not explicit. Ideally, only remote exceptions that don't
> > >have the
> > >explicit reverse (rmi-to-system) mapping should result in a ServerException.
> >
> > Right, your first point makes this not so bothersome.
> >
> > > > >2. What about C++ clients ? The server should have converted any Java
> > > > >Exception into a CORBA SYSTEM/USER exception where this is possible,
> > > so that
> > > > >any non-Java clients can deal with the Exceptions. Especially for the
> > > > >TRANSACTION exceptions where a one-to-one mapping is possible, so that C++
> > > > >clients know about those events from an RMI-IIOP server. Unfortunately the
> > > > >Table 1-2, CORBA and RMI Exceptions (p. 1-33), has no one-to-one
> > > mapping, so
> > > > >Then mapping into the other direction some information will get lost.
> > > >
> > > > The problem is actually worse than this. If the server throws an exception
> > > > that is a subclass of a declared exception, the client will get the Java
> > > > IDL mapped exception (e.g. FooException -> FooEx) but
> > > > _there_is_no_way_for_the_client_to_handle_it_ because the IDL does not
> > > > declare FooEx - and IDL does not allow exception inheritance. I would like
> > > > to propose therefore that in the "standard" exception mapping that only
> > > > exceptions of the *declared* type are thrown and that any exception
> > > > inheritance is represented in the detail which is the actual java
> > > > exception. We are actually doing this already in our server because it is
> > > > the only reasonable way to make C++ clients work with EJBs.

I'd like to raise the following as an urgent issue for the Java to IDL
RTF:

RuntimeExceptions/Errors are broken in RMI-IIOP when interoperating
between J2SE 1.3 and 1.4 (or "class evolution in service contexts is
broken").

When an unchecked exception (RuntimeException, Error) occurs in
RMI-IIOP, it is translated into a CORBA UNKNOWN system exception, and
the real Java exception is marshaled as a CDR encapsulation in the
UnknownExceptionInfo service context.  (Java to IDL formal 01-06-07
1.4.7, 1.4.8.1)

In J2SE 1.4, java.lang.Throwable has evolved to add several new
non-transient fields including

private Throwable cause;
private StackTraceElement[] stackTrace;

This means that when J2SE 1.3.x talks to 1.4 and an unchecked
exception occurs, there is a class evolution scenario occurring inside
of the UnknownExceptionInfo service context.  Normally, class
evolution is handled by using the SendingContextRunTime CodeBase to
find out what the sender put on the wire.  This service context is
sent by the client on the first request on a connection and by the
server on the first reply on a connection.  (Java to IDL formal
01-06-07 1.5.1)

This presents two problems:

1.  Since the UnknownExceptionInfo service context has the Throwable
in an encapsulation, technically it has no connection information.

2.  Even if we assumed that we use the connection on which this
encapsulation is transmitted, we would need to make sure that the
SendingContextRunTime service context is unmarshaled BEFORE the
UnknownExceptionInfo service context so that the connection has the
CodeBase to use to unmarshal the evolved Throwable class.

Here are two possible solutions:

1.  Service context ordering proposal (strawman):

Require that service context encapsulations use the connection on
which they are sent if they need the CodeBase, and that the
UnknownExceptionInfo service context must be unmarshaled after the
SendingContextRunTime service context.  Specifically, add this
language to the Java to IDL spec in section 1.4.8.1.

Note:  We may need to be careful not to generalize to all service
contexts.  For instance, code sets for wstring/wchar inside of service
context encapsulations should probably be explicitly stated when
defining the service context at the spec level.

2.  New service context & backwards compatibility proposal:

Add a new UnknownExceptionInfo service context called something like
UnknownExceptionInfoPlusCodeBase service context.  It contains the
CodeBase IOR followed by the java.lang.Throwable.  Thus, all necessary
information is contained, and we can support evolution going forward
in this particular service context.  Older receivers would just
disregard it.

For backwards compatibility with J2SE 1.3 and before, we must still
also send the original UnknownExceptionInfo service context containing
something that looks like the J2SE 1.3 java.lang.Throwable class. 
Thus, we must define the format of this new structure and how to
translate from J2SE 1.4 or later Throwables to it.

Resolution: 
In order to make progress on issue 4795, the proposal is that we split it into two parts:
a) the immediate need to fix an interoperability problem between JDK 1.3 and JDK 1.4 when certain exceptions are thrown
b) the desire expressed by many RTF members and others to allow full type information to be sent with valuetypes, so that callbacks to the SendingContextRunTime aren't needed for versioning 

This split would allow us to do a fix for part a) in this RTF to close this issue and allow JDK 1.3 to talk to JDK 1.4 correctly, and raise a new issue for part b) to be resolved by the next RTF. The fix for part b) could be part of a new GIOP level and might involve fairly significant CDR changes. 

The proposed resolution to part a) is to require that the SendingContextRunTime service context be processed before the UnknownExceptionInfo service context. If the UnknownExceptionInfo service context happens to physically precede the SendingContextRunTime service context on the wire, and no SendingContextRunTime service context had previously been processed for the connection, then the UnknownExceptionInfo service context would have to be copied to a temporary buffer and its data unmarshalled after the SendingContextRunTime service context has been processed. This should be quite easy to implement, and could be retrofitted to JDK 1.3 in a service update. 

Adopting this resolution for the immediate problem would not preclude making a further change later to add optional type information to CDR valuetypes so that the need for a callback could be eliminated.

The resolution for issue 3151 does not make it clear how to encode a custom
valuetype that has no optional custom data.  This includes custom types
whose writeObject() method calls defaultWriteObject() but writes no additional
data, and custom types that don't write any data at all.


Tihs case could be handled by always writing the org.omg.customRMI.* wrapper
even if there is no data inside of it.  However, this is very wasteful of
space.  A better approach would be to write a null valuetype (0x00000000) to
indicate the omission of the optional data.  The optional data cannot be omitted
entirely, since the receiver would not be able to correctly parse the stream.


Proposed resolution:


Change paragraph "d" in section 1.4.10 from the following:


d. (optional) Additional data written by writeObject, encoded as specified below.
   For format version 1, the data is written "as is".  For format version 2, the
   data is enclosed within a CDR custom valuetype with no codebase and repid
   "RMI:org.omg.custom.<class>" where <class> is the fully-qualified name of the
   class whose writeObject method is being invoked.


to the following:


d. Additional data written by writeObject, encoded as specified below.   For format
   version 1, this data is optional and if present must be written "as is".  For
   format version 2, if optional data is present then it must be enclosed within a
   CDR custom valuetype with no codebase and repid "RMI:org.omg.custom.<class>"
   where <class> is the fully-qualified name of the class whose writeObject method
   is being invoked.  For format version 2, if optional data is not present then a
   null valuetype (0x00000000) must be written to indicate the absence of optional data.

I'd like to raise the following as an urgent issue for the Java to IDL
RTF:


The javax.rmi.CORBA.Stub class defined in the Java to IDL
specification does not explicitly define its serialVersionUID, yet
this value is crucial for interoperability.  The sender's Stub
repository ID must be compatible with that of the receiver.


Without a standard value, even compiler differences can affect the
serialVersionUID.


In our case, we removed an unnecessary protected constructor before
our J2SE 1.4.0 implementation shipped.  Unfortunately, this broke
interoperability with our (and others') previous implementations, and
it is seriously affecting one of our customers.


We would like to standardize the following (older) value so we can
then patch our 1.4.0 implementation:


Proposed resolution:


Add the following serialVersionUID to the Stub class in Java to IDL
(ptc-02-01-12) 1.5.1.2:


// Java
public abstract class Stub
  extends org.omg.CORBA_2_3.portable.ObjectImpl
    implements java.io.Serializable {


    private static final long serialVersionUID = 1087775603798577179L;


    ...
}

.

The Java to IDL specification uses the IOR TaggedComponent
TAG_JAVA_CODEBASE to transmit a codebase string used to download stubs
and ties.  Yet, at IOR creation time when using the POA, the ORB may
not know the implementing servant, so may not be able to determine the
correct codebase.


I propose solving this by the creationg of a standard POA Policy for
the Java codebase component.  A POA using this policy will add the
given codebase string as the TAG_JAVA_CODEBASE component in the IORs
which it creates.  (Someone using a single given POA for supporting
multiple servant types would of course have to include the codebases
for all of them.)


Proposed resolution:


1. Add a section in CORBA 11.3.7:


11.3.7.x Java Codebase Policy


Objects with the JavaCodebasePolicy interface are obtained using the
ORB::create_policy operation and passed to the POA::create_POA
operation to specify the TAG_JAVA_CODEBASE TaggedComponent's value
that should be used in this POA's IORs.  The any passed to
ORB::create_policy should contain a string which is a space-separated
list of one or more URLs, as described in the Java to IDL mapping
(formal/01-06-07), Codebase Transmission.


2. Add the following to the IDL in CORBA 11.4:


const CORBA::PolicyType JAVA_CODEBASE_POLICY_ID = nn;
[note to editor: number to be assigned by OMG]


local interface JavaCodebasePolicy : CORBA::Policy {
  readonly attribute string codebase;
};

Proposed resolution:

1. Add a section in CORBA 11.3.7:

11.3.7.x Java Codebase Policy

Objects with the JavaCodebasePolicy interface are obtained using the
ORB::create_policy operation and passed to the POA::create_POA
operation to specify the TAG_JAVA_CODEBASE TaggedComponent's value
that should be used in this POA's IORs.  The any passed to
ORB::create_policy should contain a string which is a space-separated
list of one or more URLs, as described in the Java to IDL mapping
(formal/01-06-07), Codebase Transmission.

2. Add the following to the IDL in CORBA 11.4:

const CORBA::PolicyType JAVA_CODEBASE_POLICY_ID = nn;
[note to editor: number to be assigned by OMG]

local interface JavaCodebasePolicy : CORBA::Policy {
  readonly attribute string codebase;
};

1. The Java to IDL mapping spec defines the ValueHandler.getRunTimeCodeBase() that returns an CORBA object reference of SendingContext::RunTime to be used to construct the SendingContextRunTime service context:

    org.omg.SendingContext.RunTime getRunTimeCodeBase();

However, the ValueHandler is not ORB aware of,  how should the object reference be constructed by the ValueHandler? More specifically, how should the ValueHandler get a handle of a POA and which POA to create the reference?

2. Similar issue exists in PortableRemoteObject.exportObject() and PRO constructor:

    protected PortableRemoteObject() throws java.rmi.RemoteException;
    public static void exportObject(java.rmi.Remote obj) throws java.rmi.RemoteException;

The spec says: 

"Server side implementation objects may either inherit from javax.rmi.PortableRemoteObject or they may simply implement an RMI/IDL remote interface and then use the exportObject  method to register themselves as a server object"

The PRO is again not ORB aware of,  how should the remote object be exported and registered with a POA and with which POA?

3. The spec needs to clearly indicate on who owns the creation/initialization of the ORB. Should rmi-iiop implementation implicitly create/initialize an ORB instance at some point of time? Or should the RMI-IIOP application create/initialize an ORB instance and pass into the RMI-IIOP implementation via certain APIs? 

1.2.3 RMI/IDL Remote Interfaces 2. All methods in the interface are defined to throw java.rmi.RemoteException or a superclass of java.rmi.RemoteException. 

Should this read " ... or a subclass of ..." as per point 4 and elsewhere?

Resolution: Closed, no change. 
The spec wording is correct and consistent with the RMI spec. If a method of an interface declares a superclass of RemoteException (e.g., IOException) but does not declare RemoteException, then the method can throw RemoteException and all of its subclasses, as well as some other exceptions. The method therefore meets the criteria for a conforming remote interface. 

However, if the method declares a subclass of RemoteException (e.g., ConnectException) but does not declare RemoteException, then the method cannot throw RemoteException and neither can it throw other subclasses of RemoteException, such as ConnectIOException. The method therefore fails the criteria for a conforming remote interface.

You know, this scheme for mapping names is pretty byzantine. As far as I can tell, if you happen to name a java variable J_Fred, you've got problems. I'm sure you get plenty of suggestions. Here's mine: 

1) Underscore is used consistently as an escape character. Thus: _ -> __ $ -> _S \u1234 -> _U1234 ( -> _P (parenthesis) [ -> _B (bracket) ; -> _C (semiColon) / -> _F (forward slash) 

_U uppercases a single character doRecord -> DO_URECORD 

_A indicates that the entire name is in uppercase MAX_VALUE -> _AMAX__VALUE 

2) inner classes - we replace the dot with a $, and use that as the name. This is how most java compilers produce inner classes. 

3) Overloaded names - In the case of overloaded names, then rather than using the simple name of a method or variable, we use the converted name as it is defined in the java virtual machine specification, section 4.3. This is an already existing standard for uniquely identifying members in java classes. As methods are identified by parameters (ie: we don't need the return types to distinguish them), we can ignore the closing parenthesis and return type of a Method Descriptor. 

However, we use lowercase for the characters in table 4.2 so that our underscore escaping doesn't expand out foo(I to foo_P_UI 

thus: int FOO int foo void foo() Object foo(int) long foo(int[]) a.b.Boz foo(a.b.Bar.Baz) 

become 

FOO foo foo( foo(i foo([i foo(la/b/Bar$Baz; 

which are encoded by the rules above to 

_AFOO FOO FOO_P FOO_PI FOO_P_BI FOO_PLA_FA_F_UBAR_S_UBAZ_C 

These are not very natural ... but there you go. Serves you right for attempting to use IDL for overloaded names. 

If IDL had an name ailiasing facility (maybe it does), then all classes and methods could be generated as above and used as such in generated code, and an alias produced for those names which map unambiguously to a particular field or value. 

Thus for doAThing(int), all generated code would use DO_UA__UTHING_PI, but an alias DOATHING would be generated and presumably used by coders. 

Resolution: Closed, no change. 
The current spec does cause collisions, and this has been known from the outset. The design objective was to avoid collisions for normal usage of Java method and variable names, while preserving reasonable usability for IDL generated from real-world Java classes and interfaces. 

It is correct that having names _fred and J_fred within the same Java interface or class causes problems. But how many real-world Java programs do this? Is there a problem with a real Java class or interface, or is this issue based on the theoretical possibility of a clash in a contrived case? 

Some of these suggestions have merit, but at this point it is not practical to make radical changes that would affect all RMI-IIOP applications and implementations. If there is a specific example of a name clash that is causing real-world problems, then the RTF can consider making minor changes to the existing mapping to resolve these problems

Section 1.3.2.4 of the Java to IDL Mapping spec describes name mapping rules
for illegal IDL identifier characters.  The text implies that only non-Latin 1
characters are illegal, but in fact non-ASCII characters are illegal in IDL
identifiers (see section 3.2.3 of the CORBA spec).


Proposed resolution:


In section 1.3.2.4, change "outside of ISO Latin 1" to "outside of ASCII".

Questions have come up recently about copyObject behavior, both within Sun and from customers.
The main question is whether it is necessary to actually copy immutable objects, or can copyObject( obj ) 
simply return obj for immutable objects.  The standard immutable
objects in question are String and the primitive type wrappers (Integer etc.).


Looking at ptc/02-01-12, the main part of the specification that is relevant is section 1.5.1.6,
on the bottom of page 1-52.  The statement is made here that "(copyObject and copyObjects) observe
copy semantics for value objects that are equivalent to marshalling, ...".  What exactly does
equivalent mean here?


It seems to me that the most important property here is that any sequence of method invocations 
made by the "client" do not affect the results of any sequence of method invocations made by
the "server" (or vice-versa).  This is normally the case for immutable objects that are copied by reference.  
However, locking of an object passed by reference is possibly observable.  Normally, the
"client" and the "server" in the colocated case are actually running on the same thread, but
each role could pass object obj and copyObject(obj) to a different thread.  Different results 
are then possible for the copy/non-copy case, depending on the locking behavior of the threads.


There are old parts of the rmic -iiop code and the Sun ORB, dating from the days when Sun and IBM
were jointly developing RMI-IIOP, that show evidence of both possible interpretations.  For
example, StubGenerator.mustCopy( TYPE_STRING ) returns false, so that a method that takes 
only one or more String arguments would not have a call to copyObject(s) generated in the stub.
However, the implementation of Util.copyObject(s) always copies String instances.


I would like to see the text clarified in the specification to explicitly state whether immutable
objects need to be copied or not.

Resolution:
Closed, accepted. The spec already states clearly in section 1.5.1.6 that these methods do make physical copies. Clarify the spec to say that if the stub wants to optimize out the copying (which is legal), then it should omit calls to these methods. See revised text below.

We currently are having a problem implementing abstract interface behaviour in C++ for valuetypes marshaled according to the Java/IDL spec. For custom marshaled objects the spec says:


"Other Java objects are marshaled in the form of an
IDL abstract interface (i.e., a union with a boolean discriminator containing either an
object reference if the discriminator is true or a value type if the discriminator is false)."


However, the Java/IDL spec does not mandate that _all_ Java objects of this type be mapped to IDL that has an abstract interface as a supertype. Thus we are left in the position for IDL that objects marshaled as abstract interfaces are not truly abstract interfaces as defined in the CORBA spec section 6.2. Moreover the generated C++ for these types will not include inheritance from AbstractBase, so how should these types be handled? In particular should DataInputStream::read_Abstract() work at all for these types? You could imagine making this work by having read_Abstract() return some container class that held the actual object (valuetype or object reference), but its not clear that this is actual legal wrt 6.2.


Clarification appreciated.

Resolution: Closed, no change. See revised text below.

We currently are having a problem implementing abstract interface behaviour in C++ for valuetypes marshaled according to the Java/IDL spec. For custom marshaled objects the spec says:


"Other Java objects are marshaled in the form of an
IDL abstract interface (i.e., a union with a boolean discriminator containing either an
object reference if the discriminator is true or a value type if the discriminator is false)."


However, the Java/IDL spec does not mandate that _all_ Java objects of this type be mapped to IDL that has an abstract interface as a supertype. Thus we are left in the position for IDL that objects marshaled as abstract interfaces are not truly abstract interfaces as defined in the CORBA spec section 6.2. Moreover the generated C++ for these types will not include inheritance from AbstractBase, so how should these types be handled? In particular should DataInputStream::read_Abstract() work at all for these types? You could imagine making this work by having read_Abstract() return some container class that held the actual object (valuetype or object reference), but its not clear that this is actual legal wrt 6.2.


Clarification appreciated

Closed as duplicate of issue 6410.

We have recently found an interop problem with another Vendor's orb and wanted some clarification on the customer marshaling format wrt defaultWriteObject. 1.4.10 says:


For serializable objects with a writeObject method:
b. boolean - True if defaultWriteObject was called, false otherwise.


However it is mute on the subject of writeFields and readFields. The Java serialization spec indicates that defaultWriteObject and writeFields are somewhat equivalent, so (a) one could argue that the boolean should be true if writeFields has been called. (b) One could also argue that the spec doesn't say this and so the boolean should be false.  My feeling is that (a) is correct (and this is what the other orb does) but that it should be called out explicitly in the spec.


Comments appreciated

Closed, accepted. See revised text below.

The CORBA activity specification (Additional Structuring Mechanisms for
the OTS, v1.0, OMG document formal/02-09-03) introduced three new system
exceptions (INVALID_ACTIVITY, ACTIVITY_COMPLETED, and ACTIVITY_REQUIRED)
that were incorporated into CORBA 3.0.


The JCP J2EE Activity Service for Extended Transactions (JSR 95) adds
three equivalent exceptions
 javax.activity.InvalidActivityException
 javax.activity.ActivityCompletedException
 javax.activity.ActivityRequiredException
to the Java platform.


The Java to IDL mapping should be extended to map these CORBA exceptions
to the equivalent Java activity exceptions, just as the three CORBA
transaction system exceptions are currently mapped to the equivalent
JTA exceptions in the javax.transaction package.


Proposal:


In section 1.4.8, add the following rows to table 1-2:


CORBA Exception     RMI Exception
---------------     -------------
INVALID_ACTIVITY    javax.activity.InvalidActivityException
ACTIVITY_COMPLETED  javax.activity.ActivityCompletedException
ACTIVITY_REQUIRED   javax.activity.ActivityRequiredException

Closed, accepted. See revised text below

Java to IDL Language Mapping Specification doesn't cover the mapping for Java enum from JDK 1.5. We propose the following mapping to be included in the specification. Mapping for Enumeration: When used as a parameter, return type or member variable, Java enum in JDK 1.5 is mapped to the OMG IDL type ::javax::rmi::CORBA::EnumDesc. The corresponding representation in Java class is: package javax.rmi.CORBA; public class EnumDesc implements java.io.Serializable { // Holds the value of enum public String value; // Name of enum class public String className; static final long serialVersionUID = ...; } The mapping in IDL will be: module javax { module rmi { module CORBA { valuetype EnumDesc { public ::CORBA::WStringValue value; #pragma ID value "RMI:EnumDesc/value:xxxxx" public ::CORBA::WStringValue className; #pragma ID className "RMI:EnumDesc/className:xxxxx" }; #pragma ID EnumDesc "RMI:EnumDesc:aaaaaa:xxxxxxxxx" }; }; }; 

Resolution: Closed, accepted. See revised text below