quick follow-up to my previous post.

>Hello there,
>
>it seems that service invocation over an RMI binding in Tuscany 1.6 does not go
>with the best practice of loading contributions through a dedicated 
>ClassLoader,
>and not putting them on the local application classpath.
>
>This is a severe problem that I am afraid could make the RMI binding almost 
>useless.

I retract that unnecessarily strong wording. Let’s say “awkward to work with”.

[snip]

>I guess a way to make this work correctly would be for Tuscany to use its own 
>RMI registry implementation, instead of
>sun.rmi.registry.RegistryImpl. But that seems pretty radical. Another approach 
>might be to change the RMIReferenceInvoker
>so that it does not use Class.getMethod() directly but implements a modified 
>sort of reflection where the
>parameter types are compared by name and not by identity.

That second suggestion would not work, as there would not be a way to convert 
the actual parameters
from one class to the other.

I’ll have to split up my contributions and put every class that occurs as a 
method parameter on the
application classpath. I think that’s a restriction worth pointing out. It’s 
not really to do with Tuscany,
but as far as I can see just a consequence of how RMI works.


n  Sebastian
IDS Scheer Consulting GmbH
Geschäftsführer/Managing Directors: Kamyar Niroumand, Ivo Totev
Sitz/Registered office: Altenkesseler Straße 17, 66115 Saarbrücken, Germany - 
Registergericht/Commercial register: Saarbrücken HRB 19681
http://www.softwareag.com

Reply via email to