- From: <noah_mendelsohn@us.ibm.com>
- Date: Thu, 6 Feb 2003 17:07:53 -0500
- To: ht@cogsci.ed.ac.uk (Henry S. Thompson)
- Cc: richard@cogsci.ed.ac.uk, www-tag@w3.org
Henry Thompson writes: >> I made the proposal in the form I did to catch >> all and only what I understood the SOAP >> requirements to be, but I suspect you're >> right that just skipping the whole thing is better. (speaking only for myself, not officially for the XMLP WG) As far as I know, SOAP hasn't expressed a requirement for anything. SOAP is an application of XML. It so happens that a legal implementation of SOAP will never put a DOCTYPE or a PI into a SOAP message. It also won't put in an <animal:elephant> tag as a child of the <soap:envelope>; neither is allowed by SOAP. As far as I'm concerned neither restriction directly represents or suggests a requirement for anything to be included in future XML specifications. SOAP protocol bindings can use any representation they like as a wire format. Some of those may not even have representations that could correspond to a DOCTYPE, implying that there is no way a receiver could see one at all. The interesting case, of course, is when the binding chooses to use an XML 1.x serialization, which is what the supplied HTTP implementation does. In that case, you could imagine a buggy sender managing to transmit what is an otherwise legal SOAP message with a DOCTYPE or PI. With the exception of one small exception that's allowed only for performance, receivers receiving such representations must reflect errors >at the SOAP level<. It's not an XML error, it's a SOAP error. Same as if an <animal:elephant> shows up. So, nothing in this includes a "SOAP requirement" as far as I know. Some who have seen these design decisions have come to their own conclusions that a subset defined at the XML level would be better. I'm not completely convinced, but that's what the TAG has quite appropriately suggested that the XML Activity, core group and/or Advisory Committee consider per the usual W3C process. I'm glad to see that analysis starting, and I'm curious whether a subset, a conformance level, a new version of XML (deprecating the featues in question) or doing nothing will prove on balance to be the best course. Right now, I don't feel that I know the answer, but I'm quite convinced that SOAP itself has no "requirement" in this area. Thanks! ------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 IBM Corporation Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------
Received on Thursday, 6 February 2003 17:11:11 UTC