Re: IIOP tools: Working Classpath CORBA prototype released.

[Meskauskas Audrius]

Hi, Sven, Tom and Archit,

> From the four tools you listed, none is "absolutely" required "to have a 
working remoting system". However IDL compiler makes this work much easier, 
and the naming service now seems becoming a popular standard.

The development of the CORBA Internet Inter-ORB Protocol (IIOP) in our GNU 
Classpath project is not so sad as it may be seen. I have just finished and 
released the first working prototype of our ORB. It is interoperable with 
ORB, included in the Suns release 1.4, transferring the data structures and 
exceptions in both directions. After doing the beginning, I really expect 
more developers to join.

We already have the working transient naming (CosNaming) service, required 
by many CORBA applications (part of Suns jre since 1.3). When having the 
transient service, writing the persistent service becomes a clear and 
interesting task. I am sure that somebody will do this rather soon. POA is 
also not yet written - another good task for somebody to take. I think, 
these systems must belong to the Classpath core, having the Classpath 
license. If somebody wants to plug-in the alternative naming service 
instead, it is simple to do this by setting the initialisation properties.

We do not have the IDL to java compiler (idlj) yet. However I would like to 
pay your attention that IDL compiler and CORBA implementation must be fitted 
to each other. It would be an interesting work to inspect JackORB's IDL 
compiler, checking if its generated code can run with Gnu Classpath 
prototype and, if no, that kind of adaptations are required. I did not even 
checked yet how easy to separate it from the rest of JackORB. The 
development of our own IDL compiler was in my immediate future plans, but if 
we can separate and adapth the JacjORBs compiler, this may not be 
worthwhile.

At the moment, I am not implementing the other two tools, rmic and rmic-idl. 
I cannot do all at once. Here, really, it would be great if somebody could 
take this task. However the considerable part of the CORBA functionality can 
be achieved without them.

Several months ago, when all CORBA was still "0% - dark red" in the status 
review, it was time for somebody to negotiate, maybe Classpath could accept 
the CORBA implementation, having the different license than the rest of the 
project. Now, I think, this time has been passed. Working together, we can 
easily have our own CORBA, licensed exactly in the same way, as the rest of 
Classpath. Yes, CORBA seems complicated, but swing also seems complicated, 
and sockets, and io, and charsets, and many other things we have already 
done. After we have the working core and examples, the further work on CORBA 
is promising to be easier.

Regards

Audrius