Hi, > As far as I can see no one's talking about that > MIG/CORBA thing anymore. I > wouldn't mind if I knew what is happening now. Yep not much of talking now.May be more study going on now. > I'd like to aid with development, but... Is anyone > working on an > implementation of Corba for our purposes, or should > I > go on with learning about MIG? I am personally trying to get one HURD server running using CORBA. But there are some issues I am facing and working on them.But I get busy with other craps at times. It isn't piece of cake to map the current *.defs file mapping to CORBA compliant IIOP.I have been able to make/generate some of the existing code IIOP compliant i.e start with MIG IDL and produce stubs that communicate via IIOP.Although I have met with some success with some of the defs from which I could easily produce the stubs,still I am working on some other important defs for some of the HURD Servers to map the code.The problem usually comes when you try to take the *existing* '.defs' and map them.Flick has been providing quite good results for this conversions for some defs;but some of the more important defs still need a lot of flick tweaking.This is one of the things I am working on. I could find out with some of the successfully compiled defs stub code , that flick produces not just the code which is IIOP compliant but also it generates good code for the initialization of the BOA,registeration of objects for CORBA etc.But it produces this in a main() function in the stub code.so I tweaked and recompiled flick to avoid the generation of this main() in stub and rather use the main() of the existing subsystem /server in HURD.The initialization and registeration code for CORBA should now be copied in this main() server function.Now the problem that I face here is that CORBA requires registeration of Objects.So now I have to represent the server as an object,which is not a big task.Then I have to register this object.The operations on this object would be the functions defined in the server's '.defs' file. After this You have to tweak the CORBA_BOA_Create function as flick's CORBA_BOA_Create function isn't quite standard.Can't give the details here of how to do that. Then I am working on how this whole thing will fit into the flick/ORBIT runtime etc, again I can't give the details in a single mail. The bottomline is the mapping on which I am working, by flick tweaking.As flick is quite good in this we needn't start from scratch. Another interesting way is to take .IDL for new servers and then generate ipc compliant code.The problem here is IIOP compliancy and the existing .defs. Anyway working on this too. Have to look at the L4 implementation of CORBA too. Only problem is that all of this requires lot of knowledge of IPC(very big job),MIG(not a big problem)and CORBA.I lose time in between due to some diversions and crap. MIG->CORBA conversation is 'ebbing' in nature as the more deeper you get,the more you need to study before commenting any further.Moreover the initial approach may change as you go along. Again too much discussion without work/experiment leds to a conversation in completely different and varied directions and that too on assumptions.So may be people are too busy with other things or are quitely experimenting :-) to get some substantial working code. Mridul. ===== Inspired by GNU www.gnu.org __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/ _______________________________________________ Bug-hurd mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/bug-hurd
