How the "library naming scheme" works loading dynamical libraries?

[Martin Egholm Nielsen <martin at egholm-nielsen dot dk>]

Hi,

I just stumbled across:

http://gcc.gnu.org/ml/java/2002-11/msg00394.html

searching for how to implement and use gcj-natively-compiled shared 
libraries loaded dynamically (R).

In short, this posting's example loads an implementation (named "bte") 
of an interface (named "bi") dynamically at runtime using 
"Class.forName("bte")" and having LD_LIBRARY_PATH pointing to the path 
of the shared library.

Fine! However, I'm not quite familiar with the mechanism that tracks 
down the implementation, "bte", located in the library "lib-bte.so".

Does the class loader search each and every library in LD_LIBRARY_PATH 
for this implementation? In it does, when does it do - at gcj init-time 
or at Class.forName() invocation-time?

In case the library to "search" is determined by the argument to 
Class.forName(), how does one create and use a shared library with more 
than just one "implementation"?

BR,
 Martin Egholm