I recently attended the WS-ReliableMessaging Interop Summit, hosted by IBM. The agreement prohibits me from saying anything about the specific results due in part to the workshop allowing for vendors to arrive with “pre-release code”. A public version of the agreement and the test scenario document can be found here.
The way the workshop works is all the vendors sit in a room and fire messages at each other. The test scenario, as outlined in the publicly available test scenario document, is relatively simple. Its a simple ping test of sending 3 sequenced messages to another WS-ReliableMessaging enabled endpoint. The receiving endpoint responds with an acknowledgement message that contains the range of message sequence numbers that it received. That’s it. Sounds simple, but Rome wasn’t built in a day. You have to start somewhere.
Glen Daniels (now a Sonic guy) brought an implementation that he had written using Axis handlers. The question now is whether we just submit the implementation to Axis, productize it in Sonic, or both. My gut feel tells me that we should do both, and that productizing it in Sonic would yield a more robust implementation because there is already a reliable messaging infrastructure there to take advantage of. There’s a whole team of engineers back at Sonic HQ that will be happy to remind me that there’s much more to doing a reliable messaging infrastructure than a simple message exchange pattern in an Axis handler. And Glen would be the first to agree. Perhaps in the end we’ll do both.
[FYI - 10/23/2003 IBM recently published notes on the meeting]