- From: <noah_mendelsohn@us.ibm.com>
- Date: Sat, 11 May 2002 20:32:32 -0400
- To: "Anne Thomas Manes" <anne@manes.net>
- Cc: "Sam Ruby" <rubys@us.ibm.com>, www-tag@w3.org
Modelling a GET interface in WSDL is fine, but we may need to deal with it in SOAP as well. Why? I think it depends on where you think SOAP processing is to happen. First question: are there SOAP nodes at both ends, only at the requesting end, or neither? Specifically, where do you expect the SOAP processing model to apply. Let's take the response to the GET first. If it's a SOAP envelope, is there mustUnderstand checking, etc. If that's what you want, you need a SOAP binding and a message exchange pattern to specify at the SOAP level what the processing and message flow rules are. Now, consider the request. You might want to do a get from a node which is not a SOAP processor. In this case, the only SOAP message is the response. However, it's perfectly coherent to say that you want the GET itself to be a SOAP message, perhaps so that you can relay it past an intermediary, and perhaps through a second hop that need not be HTTP (this will be a moderately common scenario in corporate environemnts, IMO). In this case, you need to make sure that the HTTP GET actually represents a request at the SOAP level as well. The proposal that I've made for a REST:GET soap body element provides this. So, we need to decide which cases we want to support, and design accordingly. In many cases, a SOAP binding will be required. A WSDL HTTP convention is complementary, but does not necessarily solve the whole problem. ------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 IBM Corporation Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------
Received on Saturday, 11 May 2002 21:00:06 UTC