If I understand you correctly, are you asking for something other than reflection, runtime querying (a la XSLT) and heirarchy traversal (DOM)?
If you think of your "system" as a graph objects that can be traversed; you can require a few messages to be built into a every object, likeconstructors, toString, and generic type-conversion for type-pairs. You can even encode full type description of messages and their arguments, using an schema language XSD, exactly like SOAP does.
Gaaah, I hate underspecified problems that sound interesting.
I think I overspecified the problem. I'm looking for ways to deal with maximum ignorance, because if you can deal with ignorance of the outside that means you have achieved independence. Dependencies specification is so popular today and I want something quite the opposite. There's bloat of dependencies, but there's bloat in independence as well. How would you deal with the bloat of independence?
I think for this to be a discussion, and not, say, an exercise for me to guess what you're thinking, you would need to explain yourself better and give concrete examples.
If I understand you correctly, are you asking for something other than reflection, runtime querying (a la XSLT) and heirarchy traversal (DOM)?
If you think of your "system" as a graph objects that can be traversed; you can require a few messages to be built into a every object, likeconstructors, toString, and generic type-conversion for type-pairs. You can even encode full type description of messages and their arguments, using an schema language XSD, exactly like SOAP does.
Gaaah, I hate underspecified problems that sound interesting.