[Ikvm-developers] Limitations of IKVMC

[Kevin G. <kg...@gm...>]

First of all, let me say that this is very impressive work. 

I am interested in the prospect of being able to use the .NET visual
programming tools to write Windows GUIs on top of core Java code. How
feasible this is depends on what limitations there may be in IKVMC.

Some things that come to mind

1) We use JBoss to host EJB session beans. Has anyone tried calling a
remote EJB using IKVMC-ed code? RMI runs over sockets, so in theory
the remote calls should work. However, it also relies on passing class
files dynamically back and forth between clients and servers. If a
piece of Java code creates a class dynamically by building its byte
code, then uses a special class loader to create a new Class from this
byte code, is the IKVMC-ed code going to intercept this class
definition and convert it to a .NET class?

2) Has anyone tried using an IKVMC-ed JDBC driver (Oracle, MySQL,
whatever)? If this works, it should be fairly straightforward to
populate .NET data sets from Java ResultSets and then use the .NET
data bound controls. Am I right about that? I have done a fair amount
of Windows & COM programming, and lots of Java, so I understand the
underlying technologies fairly well but I haven't built much with
Visual Studio .NET yet.

3) I am stuck on trying to use an IKVMC-ed Apache commons logging.
Directly accessing the LogFactory and Log seems to work (though I am
not sure where the output is going)
				Log log = LogFactory.getLog("foo");
				log.info("hello");
Calling a class which is packaged in a different DLL, but makes use of
commons logging under the covers, results in a NullPointerException in
LogFactory. Are there rules I need to know about regarding how classes
can find each other across DLLs? If this is an RTFM question, just let
me know.

This project looks very promising. I don't know why I hadn't heard of
it earlier. Maybe it needs more documentation :-)

Thanks,

- kevin