- From: Jonathan Marsh <jmarsh@microsoft.com>
- Date: Tue, 20 Jun 2006 06:14:11 -0700
- To: <www-ws-desc@w3.org>
- Message-ID: <37D0366A39A9044286B2783EB4C3C4E803026FB6@RED-MSG-10.redmond.corp.microsoft.com>
I noticed from Arthur's updates to the interchange format that
BindingOperation.{http cookies} is required when the SOAP binding is
engaged. The text before that makes it sound optional (e.g. "may",
"allowed".) I think Arthur's reading is probably most nearly literally
correct, but if so, the "may" and "allowed" might need to be
strengthened a little. But I wonder if this reading is really what we
intended.
The bigger question is, whether support for the defined subset of {http
*} properties are required by all implementations of the SOAP binding or
whether the whttp:* attributes are an "optional extension" of the SOAP
binding. The latter seems a bit strange, as we don't seem to require
implementations to support a {soap underlying protocol} value of
"http://www.w3.org/2003/05/soap/bindings/HTTP/", yet everyone is
required populate the {http cookies} property, which is called out as
specifically only having meaning when used with
"http://www.w3.org/2003/05/soap/bindings/HTTP/".
Not sure what the right solution is, but it seems like we should at
least make the {http *} properties optional in the component model
unless the right {soap underlying protocol} is in use. More difficult
but possibly better would be to figure out how to treat this "nested"
extension the same as the top-level ones.
[ Jonathan Marsh ][ jmarsh@microsoft.com
<mailto:jmarsh@microsoft.com> ][ http://auburnmarshes.spaces.msn.com
]
Received on Tuesday, 20 June 2006 13:14:56 UTC