I think the loading discussion has been getting confused because we
don't have a clear set of requirements defined. In an attempt to clarify
this, here are the ones I have in mind:
* We need to be able to configure Tuscany in a test environment to
support programmatic unit and integration tests. This should not
require loading of any external configuration files.
* We need to provide a mechanism that users can use to configure
container support when programatically testing their configurations.
This should not require loading of any external configuration files.
* We need to be able to configure Tuscany from XML artifacts. These may
be files, they may be other resources such as URLs or data streams.
* We need to provide mechanisms that make it easy for extension
developers to meet all the above requirements. Developers must be
able to contribute configuration information specific to their
extension.
* Extension providers must be able to provide handlers for XML
artifacts using technologies commonly accepted and familiar
in open source environments.
* Users must be able to provide custom data values for configuration
properties and references using any Java Object.
* Users must be able to provide mechanisms that construct those Objects
from XML artifacts using a binding technology of their choice.
* The implementation should use technologies familiar to both Tuscany
developers and to other open source developers who may be interested
in contributing to the project.
* The implementation must be able to handle large configurations (1000
components) with reasonable performance and memory consumption and
linear scalability.
* The implementation must be easy to maintain long term for a typical
project participant (remembering that many participants will be
more interested in their extension than in model parsing).
Did I miss anything?
--
Jeremy
