When confronted with a question regarding the description of "subscribe-able" events in the WS-Eventing specification, I went straight to the source: Jeffrey Schlimmer. He was kind enough to answer my question (and then some):
One reason we didn't bake this into WS-Eventing is that I'm fairly sure that this is a case where one size won't fit all.
WSDL is an obvious answer, where 'notification' operations (output only) indicate events and a port implementing those operations is an event source, but there will be some simple devices that are too dumb to understand WSDL.
There could be other answers too: retrieved from a registry (like UDDI), retrieved by an explicit Get message, or even defined as a convention (add an agreed-upon reference property to an EPR).
This is an excellent question to discuss at the upcoming WS-Eventing feedback workshop; I'll suggest it to the organizers.
I remember watching Aaron do "his thing" at a CNUG meeting last year. During his talk, he demonstrated a SoapExtension that performed XML schema validation for ASMX-based service endpoints, which was very impressive. Of course, being a keen observer of all things SOAP-oriented, I asked/stated, "This is awesome stuff. Super cool. But, I have a question: Could this leverage WS-Policy?" At which point, Aaron stated, "Yes, it could." Yes, my heart skipped a beat (or two, or three).
Well, it looks like Aaron will be hitting us with that promise - via support for MessagePredicates in the form of a custom policy assertion for WSE v2.0 TP - in the next issue of MSDN Magazine:
Grab the source here.
I have a request for Microsoft: Please, please, please start a GXA/WSA/WS-* specification newsgroup/mailing list. I highly doubt
that I'm the only one who would love to discuss these specifications in an open forum.
I have a question regarding the WS-Eventing
specification: How does a [event] source describe what events a [event] sink can subscribe to? In the context of this specification, does a notification operation in WSDL
(AKA, output-only operation) imply a "subscribe-able" event?
The SPS 2003 documentation claims that this property “gets the GUID of the list to use for the lookup.” This is incorrect; it returns a string. Furthermore, this string is not guaranteed to be a GUID; it may be the name of a list. In either case, you'll have to write some pretty defensive (and rather ugly) code to get around this.
"Where's the WSE?!"
The lingering question endures...
I find myself asking this question every time I view the source of various projects hosted by GDN. In fact, I have yet to see a project - hosted by GDN - that uses WSE. Am I wrong in assuming that its adoption is low?
Many thanks to Scott for providing me with the new digs.
I will continue to post entries at my Web site but I'll be using this space exclusively for .NET-related goo.