Hi Evan,
> -----Original Message-----
> From: Evan Ireland [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, April 12, 2000 11:14 PM
> To: [EMAIL PROTECTED]
> Subject: Re: Why smoke signals matter
>
[agreed stuff deleted]
>
> > Fact 4: Because of Fact 2 and 3 you need wrappers if
> > you want access to EJBs from a language different then C++ or Java.
>
> Agreed. Actually even if you are using C++ you have a lot of work on
> your hands here. Just try using the RMI/IIOP mapping and see how much
> complex C++ code you have to write to deal with a Java 2 collection
> returned by an entity bean finder method! Do any vendors provide such
> code with their C++ CORBA 2.3 ORBs? Not that I have seen. So every
Agreed. We are currently doing it on our own.
> customer has to write it for themselves, or create wrappers even when
> using C++. Check out the white papers on the Inprise web site - they
> tell you how to write wrappers to access EJBs in their own EJB server
> when using their C++ CORBA 2.3 ORB that supports OBV!
>
> You assume that just because an OBV mapping exists for C++ that
> somehow EJB->C++ interoperability comes for free. It does not.
I do not assume it comes for free. We are doing the work right now. Seee
above.
> If someone would be kind enough to write a C++ implementation of the
> valuetypes for Java 2 collections and make it open source, that would
> do the EJB/CORBA community a great service at least in respect of
> EJB->C++ interoperability. It would be great for a university research
> project. Until such time, every single customer trying to achieve
> EJB->C++ interoperability using the RMI/IIOP mapping has to reinvent
> the wheel! This is not far off being totally ridiculous!
This would be great and I think sooner or later this will happen.
Hey your EJB vendors. Sponsor some Open-Source project doing exactly
this ! Mybe some folks from MICO or TAO would be interested !?
> > So the conclusion for the problem should be:
> > Extend the language mappings that currently lack valuetype
> > support to support valuetypes.
> > Then you can implement a COBOL ORB with valuetype support for your
> > mainframe.
>
> And when will that be ready? 2 or 3 years from now perhaps.
> Our customers
> demand and expect interoperability from third party CORBA 2.0
> ORBs with
> EJB components TODAY, not years from now.
Agreed. Because we are such a customer ;-)
OK. We can use C++ and are on our way to implement the missing
valuetypes. They will be small wrappers around an existing
C++ library (exactly what the Inprise paper suggest).
We go even one step further and combine this with a COM wrapper
so we can access our EJBs with VisualBasic or Delphi.
That's another point where work is needed.
OMG: Extend the COM/CORBA Interop spec to support valuetypes !
> > What you propose is:
> > - Restrict yourself to CORBA 2.0.
> > - Redefine the RMI/IIOP spec to be based on 2.0 features.
>
> No, I am suggesting defining an alternative second mapping,
> not replacing
> the existing mapping which is obviously useful for Java to Java
> interoperability.
>
> > If you follow this way you block all innovations from/to CORBA.
>
> Not at all. Please understand that I am suggesting an additional
> mapping that is more suitable for interoperability across language
> boundaries. Let the customers choose the mapping that best serves
> their interoperability requirements!
OK. Sorry to misunderstand you. I thought you want to replace the
current mapping.
Mmh, not sure about the problems with having 2 mappings!
I will need some time to think about it.
With kind regards
Oliver Geisser
Development/System Architect Fon: +49 (0)521 9318-324
CE Computer Equipment AG Fax: +49 (0)521 9318-444
Herforder Stra�e 155a E-Mail: [EMAIL PROTECTED]
33609 Bielefeld http://www.ce-ag.com
Germany
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
