- From: David Hull <dmh@tibco.com>
- Date: Tue, 04 Apr 2006 17:30:09 -0400
- To: noah_mendelsohn@us.ibm.com
- Cc: Mark Baker <distobj@acm.org>, "michael.mahan@nokia.com" <michael.mahan@nokia.com>, xml-dist-app@w3.org
> >I believe that in this respect, the reliable messaging folks would be >stretching HTTP of an RM-related 202+envelope were sent as a response to a >request that contained no RM header. As someone (Chris?) observed last >week, most requests in this scenario do contain an RM header, so it's at >least plausible to suggest that any RM-related response was at least in >some sense solicited, and is thus OK. What are your thoughts on this? >Thanks. > > The interesting point here is that such a message exists in different (but related) correlation contexts. On the one hand, it exists as part of an RM sequence. On the other hand, it exists as part of an HTTP request-response exchange. It may also exist as part of an application-level request-response or similar exchange. If I understand correctly, the spirit of 202 is that any response should be in the context of the HTTP exchange (and not something larger). For example, if I submitted a request to update my profile, I might get back a 200 along with an application-level response, telling me that my profile has been updated, or I might get back a 202 telling me that my update request arrived but not that it was processed, or perhaps even a 202 with a detailed explanation of what processing has been done and when completion is expected, but I wouldn't expect to get back (say) 202 with a statement of activity on my account for the last three months. WS-RX is, I think, implicitly widening the context to the RM sequence. Or another way to look at it is that the message carries two payloads, one an RM header which basically means "This message is part of sequence X. Please let me know what you've received so far on sequence X" and the other the application-level payload. I'm not endorsing either this description or the behavior. My gut feeling is that RX is in danger of trying to have its cake and eat it too. However, the RX-level conversation really is a perfectly good request-response. By itself it wouldn't be a problem (and I believe RX supports just such a stand-alone operation, if only as an empty message with "acknowledgment required" true). Turning it around, one could view the question as whether RX should be allowed to piggyback an application-level payload on top of a request for acknowledgment status. This seems somehow twisted to me, but interesting nonetheless. >Noah > >-------------------------------------- >Noah Mendelsohn >IBM Corporation >One Rogers Street >Cambridge, MA 02142 >1-617-693-4036 >-------------------------------------- > > > > > > > >
Received on Tuesday, 4 April 2006 21:30:38 UTC