Re: ORBit2 questions - TODO

[Michael Meeks <michael ximian com>]

Hi Doug,

On Thu, 24 May 2001, Douglas S. Keester wrote:
> 1) Is there a set TODO list for ORBit2?
 
        Firstly you need to use the orbit-small branch of ORBit2 to get   
anything to work at all :-) - otherwise the TODO will seem insurmountable.

        A TODO might look like this: 

                1. Port DynAny from ORBit-stable to ORBit2

                2. Continue profiling ORBit2

                3. Virtualize the small stub / skel invocation mechanism
                to help language bindings

                4. Add a 'super small' entry point for object creation  
                that does not go via the Object key [ for bonobo, ie.  
                an object key would only be created when the object was
                remoted ], see libbonobo/bonobo-object for the kind of
                creation semantics desired.

                5. Add an 'strace' like capability, this would be
                pluggable per object, and sit in front of the remoting    
                API. It would be macro generated for base types [ the
                macros and printf formats are already in orbit-small.c ].

                6. Fix the POAObject leaks - see
                corba-object.c (CORBA_Object_release_cb), Elliot is       
                hopefully working on this.
  
                7. Accelerate and improve type encoding by using a
                typecode hash instead of GSList *prior_tcs.
  
                8. Finish profile tag comparisons in
                corba-object.c (IOP_Profile_equal), we need to match
                IOP_TAG_MULTIPLE_COMPONENTS - a simple, but long winded
                task. It would be nice also to improve the (profile_hash)
                function to handle MULTIPLE_COMPONENTS more sensibly,
                particularly the IOP_TAG_COMPLETE_OBJECT_KEY, which is
                used by the GIOP tag.
 
        So, I'm planning to work on 1 and 3 in the short term - 1. should
have been complete, but I was delayed by recursive type marshaling bugs [
now  fixed, about to commit ].

        The _really_ fun one, that I'm happy to leave to some unpaid
person who wants glory :-) is no. 5. in my opinion. Not only is it not
difficult [ although you should probably consult
ORBit2/docs/internals/internal-indirection.gnumeric while hacking ], but
it will have a beautiful and most helpful effect for a lot of people -
being able to see each CORBA method nicely printed with its arguments.

> 2) What version of the CORBA spec is ORBit2 going to comply with?
> Still 2.2? Higher?

        The latest and greatest I think. Sadly the C binding spec is not
keeping up with the spec AFAIK, either way - if you find anything lagging
hack on it.

        I look forward to patches flooding in from all and sundry :-)
several of the above issues are somewhat trivial, and just a matter of
good testing and carefuly implementation.

        Regards,

                Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot