Well, it's been an interesting couple of weeks in the SOA Corona camp. After starting with a complete hosting of Axis 1.3 under the Equinox HTTP service, I backed off due to the tight coupling of Axis and XML beans, and implemented a (very partial) SOAP stack using EMF's ecore as a serialization format. That seemed to work relatively well - well enough to complete our target POC for EclipseCon - but now I'm revisting the situation, in part due to the fact that I'd like to be able to host existing Axis services within Corona, and partly because it seems silly to completely reinvent the axis wheel.

I'd begun task of rehosting Axis as the basis for an eclipse SOAP core, and applying a very modest refactoring to support the plugability of an ecore serialization strategy along with a mapping of Axis' commons.discovery mechanism into the eclipse plugin.xml mechanism. That was going quite well, I'd completed the implementation of a provider extension point and was moving on towards a WSDD-oriented extension point, when I made the fatefull mistake of looking at Axis 2.0.

In the new version of Axis, the WSDD is gone, replaced by server, module and service XML files. The old Axis architecture has been revamped to allow a more modular and pluggable environment to be created, primarily to ease the integration of WS-* implementations as they roll off the line. For the long run (as in a Corona release 1.0 target), supporting Axis 2.0 as the basis for a SOAP core looks like the preferable option. Folding this into our Corona prototype prior to EclipseCon feels a bit like a stretch goal, however.

I'll continue my evaluation and post more updates shortly. If anyone has any insights or opinions of how they'd like to see the new Axis architecture hosted under Eclipse, please feel free to post!

Cheers,
Joel Hawkins