William Vambenepe published his XMLFrag header proposal earlier this week and it immediately caught my eye. This was in part because of the idea of a ‘WS-SubversiveAddressing’ specification (WS-Addressing has been a lot of my day job at IBM for coming up on 2 years), and in part because apparently that title came from a colleague and I’ve been trying to work out who it was.
I’ve been back and forth on the proposal and I think there’s something with potential in there. After some thought it began to make some sense when I started to think of the XMLFrag header as a ContentType header for WS-Transfer (the W3C TAG apparently had a problem with WS-Transfer because it lacks feature compared to HTTP, but I digress). It lets the client describe to the server the representation it would like returned. That, for WS-Transfer at least, makes sense to me. Whether it’s a good idea for any other SOAP Web Service is less clear, but on the balance I suspect it might be useful.
Where I agree less it with the usefulness of embedding the header in an EPR. The use of it as an equivalent of ContentType means that it’s describing an attribute of the client, not the endpoint, which means it’s really not information which should be in the EPR. On top of that, it’s not entirely clear to me how whoever creates the EPR knows what subset of information the client wants. In general, if you’re a client using an EPR as a [destination] you got it from somewhere else, and how would that elsewhere know what you, as a client wants? Finally, an EPR represents and endpoint which probably has multiple operation associated with it, and the XMLFrag probably doesn’t make sense for more that one of those operations so makes an EPR of limited use.
So I don’t think the EPR related part is of much use, and has some conceptual problems, however I am a realist. The WS-Addressing spec allows anyone to use ReferenceParameters to generate SOAP Headers and people will use that flexibility to put anything in there, advisable or not. This flexibility was a source of much disagreement in the WS-A Working Group before I joined, and I think we still need time to tell if it was the right call.
So, the XMLFrag header looks to me to be potentially useful, and the EPR related bits not so much.
Subversive? Not to me.
