When creating custom serializers to handle custom types, I need to
specify the serializer classes in muse.xml on the server side. Muse
will automatically load and handle them for me. But, on the client
side, my code must still explicitly load and register the serializers
before it can handle the custom types.
This brings up an idea: most of the current Muse implementation focuses
on the server side framework. I think a more robust client framework
should be built, too. For example, since the server framework can
autoload from wsdls and descriptor files, the same logic can be expanded
on the client side. This will:
1) Shield client developers from needing to know the low level details
of the server design, like what serializers to use for what custom
types, the xml structure of requests/responses, etc. Clients shouldn't
need to deal with serializers directly. The framework should do that
automatically.
2) Reduce the code that client developers need to write just to make a
single call to the server.
3) Allow clients to build on a framework consisting of config/descriptor
files just as is done on the server.
I'm guessing this effort shouldn't take much work because the API's are
already there. They just need to be modularized and exposed to the
client, or built into a client framework. Any thoughts?
-Vinh
